> Given the TOC Augmented Traveler Nantucket.md file, make sure that all the other files are referenced by the TOC. Check the order for logical progression of general to specific. Adjust the TOC as necessary. --- I've updated the TOC. It now links every project doc, plus the two PDF exports. **What was missing:** - `Conversation.md` wasn't in the TOC. It's now in an appendix as the working record of the research thread. - The two PDFs (Opportunities and Ground Rules) weren't listed either. They're in the same appendix. **What moved:** - **Specification and Pipeline** were under "General Digital Humans," but both are about historical characters. They now come after the case studies and ground rules. - **Ground Rules** now come before the Specification, so the principles are set out before the production details. - **Nantucket Common Pipeline** now follows the Tier 1 overview and comes before the individual figure briefs. It's the shared machine behind all five figures, so the broader note leads. - The old "Additional Experiences" section is now two sections: - **Season One — Maria Mitchell**: the corpus inventory and the MMA/NHA ask. - **Visitor Experiences**: island-wide AR/VR first, then the single Vestal House room. I also added short labels to a few entries that weren't clear from the title alone. **Two things you may want to fix:** - **`Pipeline for a full-body 3D Virtual Human.md` is only the original prompt, with no answer.** The Specification holds the actual pipeline. You could delete this note or merge it into the Specification. - **Two notes disagree about the 1847 comet.** `Nantucket Island AR and VR` sets the Vestal roof VR scene on 1 Oct 1847. `Vestal House VR Room` and its citations say the family had sold the Vestal house by 1837, and that she found the comet from the Pacific National Bank roof. The AR/VR note needs its date or location corrected. --- > I've put each item on its own line and uploaded it to the directory. > Can you create a comprehensive pitch document with this same logical progression? Give me bare outline first, then we can create a PDF. I am going to have the .md directory published as a website, so that changes can be made. find it at: > https://publish.obsidian.md/jmarshallpittman/Augmented+Traveler/Augmented+Traveler+Nantucket+TOC --- Here's the bare outline. Sections 1–9 follow the TOC's order, from the broad field down to specific visitor experiences. A cover page, an opening summary and a closing ask frame them as a pitch. In the PDF, each section will end with a "Read more" link to its note on your published site. **Augmented Traveler Nantucket — Pitch** 0. **Cover and Executive Summary** - One-sentence premise - What we're asking for, from whom, at what budget - Link to the published site 1. **The Field: Digital Humans in 2026** - Market size and the paying verticals - Three production modes (neural video, stylized, photoreal 3D) - Lesson from Soul Machines: buyers pay for a job done, not "a soul" 2. **Historical Virtual Characters** - Four markets (museum installs, testimony, chat apps, stage holograms) - What survives: a closed set of sources, a real venue, a scholar - Cost bands by type 3. **Proven Models** - Bonjour Vincent (letters, a historian, a museum kiosk) - Dimensions in Testimony / StoryFile (recorded answers played back, not generated) - What we take from each 4. **Ground Rules** - Tell the public what it is (an interpretation, a cutoff year, credits) - Three separate source sets (own words, contemporaries, period knowledge) - The three-part Chautauqua program (monologue, questions in character, then out of character) - Things the character must never do; the three tests before anything ships 5. **The Product: One Character, Three Surfaces** - Kiosk / AR / VR architecture - Eight-stage pipeline - Kinds of events (monologue, reenactment, object lesson, two-character programs) - Team, timeline, budget bands 6. **Why Nantucket** - Archive depth (NHA, MMA, Vassar, Atheneum) and rooms still standing - Figures that pass the "enough of their own words" test - What we won't build (Wampanoag first-person, a talking Douglass, the Petticoat Row myth) 7. **The Tier 1 Slate** - Mitchell, the Essex cluster, Mott, Absalom Boston, Veeder and Brock (one paragraph each) - Shared studio pipeline: scholars, legal, art, AI, curation, maintenance - Partner map 8. **Season One: Maria Mitchell** - Why she goes first - Source inventory and what's ready now - The MMA/NHA ask: what we need from each, what we won't do 9. **Visitor Experiences** - Island-wide AR: five walking loops, deliberately modest in year one - VR rooms, with the Vestal House room design - A tourist's day 10. **Budget, Timeline and Build Order** - Season 1 → 1b → 2 → later - $150–250k for Season 1; later figures cost only sources, wardrobe and venue 11. **The Ask and Next Steps** - First-meeting agenda - Contacts **Appendix:** sources (linked to the citation notes), methodology page, credits. The PDF will need one fix first: the island AR/VR note still puts the 1847 comet on the Vestal roof. I'll use the corrected account from the Vestal House note (the Pacific National Bank roof). Before I draft, who is this pitch for? MMA/NHA, a funder, or a general audience decides how hard sections 1–5 sell versus explain. --- > This pitch is for a head-hunter that can help us to secure funds. They are non-technical, but familiar with Nantucket. We might leave room for additional historical figures and experiences, so the pitch is for a studio, not a specific character. --- Here's the outline revised for a non-technical fundraiser who knows Nantucket, pitching a studio rather than one character. It still runs from general to specific. Maria Mitchell becomes the first production and proof of concept, and there's an open section for figures and experiences we haven't picked yet. **Augmented Traveler Nantucket — Studio Pitch** 0. **Cover and One-Page Summary** - What the studio is, in one sentence - The opportunity, the first production, the funding sought - Link to the published site for depth 1. **The Idea in Plain Terms** - A life-size historical person you can meet at a museum kiosk, see through your phone on Main Street, or visit in a headset - Built from their own letters and diaries, not invented - One short scene as a picture of the visitor experience 2. **Why Now** - The technology has become affordable - Museums and destinations are already doing this (Van Gogh at the Musée d'Orsay, Dalí, Holocaust survivor testimony) - What failed and why (costly firms with no clear job; chatbots that make things up) 3. **Why Nantucket** - Unusually deep archives, and the buildings are still standing - A steady flow of summer visitors, plus school and member audiences - Existing institutions to partner with (NHA, MMA, Atheneum, MAAH) 4. **What the Studio Is** - A repeatable production process shared across every character - One character works in three places: kiosk, phone, headset - Who's on the team: historians, legal, artists, animators, AI, curation, upkeep - Each new character costs less than the one before 5. **Our Standards (Why Institutions Will Trust Us)** - Always tell visitors it's an interpretation, with a cutoff year and credits - Sources come first, and a historian can veto anything - A clear list of things we won't do - Why this matters to funders: lower reputational risk 6. **The Slate** - First production: Maria Mitchell - In development: the Essex survivors, Lucretia Mott, Absalom Boston and the Meeting House, the whaling-wives' journals - **Open slots:** how we evaluate future figures and experiences, with room to grow 7. **First Production: Maria Mitchell** - Why she goes first - What visitors get: a kiosk conversation, a Vestal House VR room, and a Mitchell AR stop at Vestal Street, the Atheneum and the Pacific Bank - Partners and what we're asking of them 8. **Experiences Across the Island** - Five walking and biking loops for a phone - VR rooms in partner venues - A visitor's day on Nantucket 9. **Business Model** - Venue commissions, licensing, school programs, the tourist app, grants - Where revenue repeats and where costs repeat 10. **Roadmap** - Season 1 → 1b → 2 → later, with new figures added from the open slots 11. **Funding Sought and Use of Funds** - First production: about $150–250k - Studio setup and runway: **to be filled in by you** - Milestones funders can track 12. **Team and Contacts** **Appendix:** a one-page glossary for non-technical readers, and "Read more" links to the published notes. Two things only you can supply: the studio's total funding target (the research only covers costs per character) and the team or founder details for section 12. I can leave clearly marked placeholders for both if you'd rather fill them in later. Does this order work before I build the PDF? --- > This outline looks fine and we can fill in details after review. Proceed. --- I've drafted the studio pitch as a 15-page PDF, written in plain language for a non-technical reader. It's in your AugmentedTravelerNantucket folder, along with a matching .md file you can publish and edit with the other notes. I also added the .md to the project. **Links to your site:** each section ends with a "Read more" box linking to the matching note on your published site. I built those links from your TOC's web address, so please click a few to check they open. **Maria Mitchell and the comet:** the VR room is set in the 1830s house, with a labeled door to the Pacific National Bank roof, where she found the 1847 comet. **Still to fill in (marked in brackets):** - The total amount you're raising and how long it's meant to last. - Your expected income and pricing, once partners agree terms. - The team, founder and advisor details. The year-one estimate of $230k–$400k simply adds the shared studio setup cost to Mitchell's production cost. Once you've marked up the .md, I can rebuild the PDF from it. --- > This color scheme does not have a readable contrast. Please do not use dark full-page background colors. Stick to darker text, with a much lighter version for a full page of color. The studio is called Augmented Traveler and the project would be Augmented Traveler Nantucket. --- I've rebuilt the pitch with dark text on light backgrounds and updated both the PDF and the .md in your folder. - **Colors:** the cover now uses a pale blue-grey background with dark navy text. Table headers are also light with dark text instead of white on navy. - **Names:** the cover title is now **Augmented Traveler**, with "First project: Augmented Traveler Nantucket" below it. The Summary names Augmented Traveler as the studio and Augmented Traveler Nantucket as its first project. The page footers now read "Augmented Traveler · Studio Pitch." The file name still reads "Augmented Traveler Nantucket — Studio Pitch" so no duplicate copy appears in your folder; I can rename it to "Augmented Traveler — Studio Pitch" if you'd prefer. --- > on the cover, avoid orphans ("lived.") and you can end the first sub-heading on the "meet;" and the second line on the "words," --- The cover now breaks after "meet:" and "words," so "at the places they lived." is a full third line instead of an orphan. I kept the colon after "meet" because the sentence already had one; tell me if you wanted a semicolon there instead. The updated PDF is in your folder. I also fixed the bottom label, which had split with a stray "·" at the end of its first line. It now breaks cleanly after "Augmented Traveler Nantucket." --- > I need to update the note on Frederick Douglass, since this is a possible project, include this as in the .md file attached: an experience, not a mind. Douglass can be a project in another place as a virtual historic character. I've added his possible project and two others in the TOC under 10.Brainstorm projects. Update the Studio Pitch where necessary. --- I've updated the Douglass note and the studio pitch. Both are in your folder, and I also saved the note and your new TOC to the project. **Douglass note:** - **Experience, not a mind:** it now states this near the top. On Nantucket he is one evening in one room. - **New section, "Douglass beyond Nantucket":** a full Douglass character is a separate project for another place. It lists New Bedford, Rochester, Cedar Hill in Washington, and Talbot County/Baltimore as possible homes. Each version of him gets its own label and cutoff year, and the 1841 night stays a scene. **Studio pitch:** - **Section 3:** it now rules out an all-purpose "talking Douglass" and points to the 1841 experience instead. - **New section 6b, "Brainstorm projects":** - The 1841 Atheneum night: one performance, a Garrison coda played by a second actor, the segregated steamer and the Atheneum lockout, its partners, and $80k–$160k. - Douglass beyond Nantucket, framed as proof that the studio can take its methods to other destinations. - Ahab and Starbuck aboard the _Pequod_. - The returning whaleship. - **Roadmap and funding table:** both now include the Douglass experience. - **Links:** section 6b links to the four new notes, and the appendix links to "Citations — Brainstorm Projects." The Ahab/Starbuck and Returning Whaleship entries are one line each with a "details to come" marker. Only the Douglass note was attached, and neither of the other two is in your folder or the project. If you attach those two notes, I'll fill in both entries. --- > here are the other two. --- I've replaced both "details to come" entries in section 6b with full descriptions from your notes. The updated PDF and .md are in your folder, and both notes are now saved in the project. - **Ahab and Starbuck aboard the _Pequod_:** described as a labeled literary scene built on Melville's 1851 text, not history. Two actors perform on a simplified deck, and the placard is honest about the novel. It sits in the Whaling Museum beside the real captains. Estimated at $60k–$140k. - **The returning whaleship:** built around a documented landing: the _Nauticon_, the _Lexington_, or the _Joseph Starbuck_. Visitors see oil casks and whalebone come ashore and learn how crews were paid a share of the voyage. China goods are labeled as souvenirs, and only a harbor clerk answers questions. Estimated at $70k–$160k. Your note sets the _Pequod_ scene in season 2, so I added it there in the roadmap. The returning whaleship has no season in its note, so I put it under "Later." Both are also in the funding table now. --- > In 10. Roadmap, can we list each experience on its own line in the table, instead of delimiting with semicolons. This would make it clear that there are multiple project in each season. --- Good; other tables unaffected. Delivering.The roadmap now puts each experience on its own row, with the season name only on the first row of its group. A navy line separates the seasons, so each one reads as a block of several projects. The updated PDF and .md are in your folder. I also split Maria Mitchell's Season 1 work into its three pieces: the kiosk, the Vestal House VR room and the phone postcards. If you'd rather count her as one production, I can put them back on a single row. --- > So, I have update the TOC on the website, you may want to update the links in the Pitch PDF to reflect the new folder structure: > > https://publish.obsidian.md/jmarshallpittman/Augmented+Traveler/Augmented+Traveler+Nantucket+TOC --- Let me spot-check that one of these links opens a real page.All 30 links in the pitch PDF now point to your new folders: Pipeline, Pitch and Nantucket, with the TOC at the top. I opened the Maria Mitchell link in the browser and it loads the right page. The updated PDF is in your folder. The .md version didn't need changing, since its note-name links work on the site after the move. Two things on your site: - **Duplicate pitch PDF.** Your published site has two copies of the pitch PDF: one in the Pitch folder and an older dated one (`2026-09-19-a-…`) in the Nantucket folder. You'll probably want to unpublish the older one. - **Pipeline note isn't published.** "Pipeline for a full-body 3D Virtual Human" isn't on the site. That's fine for the pitch, which doesn't link to it. --- > Research the pipeline for creating full-body 3d Virtual historical human characters, using AI minds, Maya, MetaHuman, Unreal (or Unity), and any associated hardware needed for development, authoring, and deployment via kiosk, mobile AR (android and apple), and VR headset (quest, vision pro, vive) in the current market. compare with the document I just uploaded. > > Give me a map of the pipeline, then identify specific roles that need to be hired, and then specific technical skills. Later, we can identify training and tutorials for the process. > > I'd like to create a directory structure and store the results in .md markdown files for Obisidian, including directories for process outlines and tutorials like: pipeline/ maya/ unreal/ metahuman/ unity/ quest/ appleVisionPro/ iPhone/ androidPhone/ etc. so that files are stored in the Claude project and eventually viewed in obsidian. --- I've created a Production folder of 37 linked Obsidian notes. It's saved in the Claude project and in your AugmentedTravelerNantucket folder, and the Production Index lists everything. **The pipeline:** 1. **Gate:** decide whether a figure is worth building and whether the rights are clear. 2. **Research:** the historian builds the source library and question bank. 3. **Likeness:** sculpt in Maya or ZBrush, convert with MetaHuman's mesh tool, then add hair and period clothing. 4. **Voice:** a credited actor records a voice bible, which becomes a locked synthetic voice. 5. **Motion:** MetaHuman's capture tools record face and body; Maya cleans up the takes. 6. **Mind:** the conversation system looks up answers in the sources first, with a rules layer enforcing the cutoff year. 7. **Events:** monologue, in-character questions, then out of character. 8. **Build:** Unreal 5.8 packages one project for the kiosk, the headsets (Quest, Vive, Android XR), Vision Pro and phones. 9. **Test:** a second historian, a class, accessibility review and deliberately difficult questions. 10. **Operate:** staff review visitor questions, the sources are versioned, and headsets are managed centrally. The Pipeline Map note draws this as a diagram and adds a table of tools, hardware and owners for each stage. **How the market compares with your Specification.** The core design holds up: one character package, Unreal and MetaHuman for the body, look-up-first conversation, three delivery surfaces. What has changed: - **Engine:** Unreal 5.8 (June 2026) is the last UE5 release before UE6, so lock each season to one engine version. - **Cheaper capture:** MetaHuman can now record full-body motion from a single camera, and its mesh tool converts any sculpted head _and_ body into a rigged character. - **Crowds:** MetaHuman crowd tools make filled halls possible, such as the Douglass 1841 Atheneum or the whaleship dock. - **Lip sync:** NVIDIA Audio2Face is now open source, so it no longer means paying a vendor. - **Quest:** Meta stopped selling business versions of the Quest in February 2026. Buy consumer units and use a third-party device manager (ArborXR or ManageXR). - **Browser AR:** 8th Wall shut down in February 2026. Prefer app-free 3D models opened from a QR code on each phone. **What the Specification was missing:** - What Maya actually does in the pipeline. - An Apple silicon Mac, which iPhone and Vision Pro builds require. - MetaHuman's body capture runs only on Windows. - Vision Pro: Unreal supports full immersion only, and the market is small. - Vive Focus Vision as the enterprise headset option. - Android XR, which is worth watching but not building for yet. - Holographic kiosk cabinets. I recommend Unreal as the main engine. Unity is worth using only for Vision Pro mixed reality or phone AR, because moving MetaHumans into Unity still breaks skin and hair. **Roles.** Seven core hires: 1. Executive producer 2. Technical director 3. Character artist 4. Unreal engineer 5. AI/conversation engineer 6. Conversation designer 7. Technical animator Contract the rest per production: a historian for each figure, a corpus editor, hair and costume artists, an environment artist, XR and mobile AR engineers, a mocap studio, a voice director and actors, a sound designer, a kiosk installer, an accessibility designer and QA. The hiring plan note gives a month-by-month order. It also covers roles the Specification left out: technical director, DevOps, a Vision Pro developer and someone running headsets and kiosks day to day. **Skills.** The skills note lists specific skills by role, written so they can be used in job descriptions. **Folders:** - `pipeline/`, `roles/` and `skills/` hold the map, the Specification comparison, hardware, citations, the hiring plan and the skills list. - `maya/`, `metahuman/`, `unreal/`, `unity/`, `ai-mind/`, `voice/`, `mocap/`, `kiosk/`, `quest/`, `appleVisionPro/`, `vive/`, `androidXR/`, `iPhone/`, `androidPhone/` and `webAR/` each have a Process note and an empty Tutorials note, ready for the training pass. These notes aren't in your TOC yet. Add the Production Index there and the rest can be reached from it. Sources: - [Unreal Engine 5.8 release](https://www.unrealengine.com/news/unreal-engine-5-8-is-now-available) - [MetaHuman 5.8 release notes](https://dev.epicgames.com/documentation/metahuman/metahuman-5-8-release-notes-in-unreal-engine) - [MetaHuman 5.7 release](https://www.metahuman.com/releases/metahuman-5-7-is-now-available) - [MetaHuman usable in Unity and Godot (CG Channel)](https://www.cgchannel.com/2025/06/you-can-now-sell-metahumans-or-use-them-in-unity-or-godot/) - [MetaHuman to Unity HDRP discussion](https://discussions.unity.com/t/metahuman-to-unity-hdrp-best-workflow/1713711) - [MetaHuman real-time animation](https://dev.epicgames.com/documentation/metahuman/realtime-animation-for-metahumans-in-unreal-engine) - [Pixel Streaming 2](https://dev.epicgames.com/documentation/en-us/unreal-engine/pixel-streaming-2-overview-in-unreal-engine) - [Maya 2027 (CG Channel)](https://www.cgchannel.com/2026/03/autodesk-releases-maya-2027-and-maya-creative-2027/) - [Unity 6.3 LTS](https://unity.com/blog/unity-6-3-lts-is-now-available) - [PolySpatial visionOS requirements](https://docs.unity3d.com/Packages/[email protected]/manual/Requirements.html) - [Meta Quest changes Jan 2026 (Android Central)](https://www.androidcentral.com/gaming/virtual-reality/meta-quest-horizon-worlds-and-game-development-changes-jan-2026) - [Meta HMS changes (ArborXR)](https://arborxr.com/blog/a-practical-guide-to-the-meta-hms-changes-for-xr-teams) - [Apple Vision Pro M5](https://www.apple.com/newsroom/2025/10/apple-vision-pro-upgraded-with-the-m5-chip-and-dual-knit-band/) - [visionOS 26 review (AppleInsider)](https://appleinsider.com/articles/26/05/29/the-last-visionos-26-review-apathy-about-apple-vision-pro-on-display) - [Unreal visionOS full immersion (80.lv)](https://80.lv/articles/unreal-engine-will-get-support-for-apple-vision-pro-s-full-immersion) - [Vive Focus Vision](https://business.vive.com/us/product/vive-focus-vision/) - [Android XR engine updates](https://android-developers.googleblog.com/2026/05/android-xr-updates-unity-unreal-godot.html) - [Android XR mid-2026 (VR.org)](https://vr.org/articles/android-xr-mid-year-2026-report-card) - [8th Wall shutdown (Kivicube)](https://www.kivicube.com/post/augmented-reality-after-8th-wall-shutdown-your-2026-guide-to-the-right-webar-platform/) - [ARCore Geospatial](https://developers.google.com/ar/develop/geospatial) - [NVIDIA Audio2Face-3D](https://github.com/NVIDIA/Audio2Face-3D) - [Convai Unreal SDK](https://github.com/Conv-AI/Convai-UnrealEngine-SDK) - [Proto Luma](https://protohologram.com/the-luma/) --- > can you generate a citations/ directory and put the sources links on a dated page? > then I'd like a hardware specifications list with currently available or near-release hardward we need for development. then, hardware needed for deployment. a price list and budget with an acquisition as-needed schedule would be nice. then we can start estimating budget over the first year of projects. --- > actually, can we revise the directory structure? I was looking at Production Index and saw Start Here, Tools, and Delivery Surfaces, which should all be directory-level organization instead of the flat directories under Production/ --- Links check out. Now I'll restructure your local folder in place.Folder moved. Now committing the 10 new or changed notes.Your local folder is done. Next I'll mirror the new structure in the Claude project.Now the tool and delivery-surface notes at their new paths.New paths are in. Removing the 36 old flat copies from the project so nothing is duplicated.I've reorganized the Production folder and added the dated sources page, the hardware notes and a first-year budget draft. Your local folder and the Claude project now have the same structure, and all links still work. **New folder structure:** the Production Index now points into these instead of one flat list. - `01 Start Here/`: pipeline, roles, skills, hardware and budget. - `02 Tools/`: Maya, MetaHuman, Unreal, Unity, AI mind, voice and mocap. - `03 Delivery Surfaces/`: kiosk, Quest, Vision Pro, Vive, Android XR, iPhone, Android phone and WebAR. - `citations/`: a small index plus **Sources — 2026-09-19**, which holds the earlier research links and today's price sources. Future research sessions each get their own dated page. **Hardware prices have gone up sharply this year.** AI data centers are buying up memory, which has pushed consumer prices far above launch: - **RTX 5090:** about $4,300 in-store and $6,400+ online. It launched at $1,999. - **RTX 5080:** about $1,600. - **128 GB of memory:** about $2,339. - **Quest 3:** $600. - **Vision Pro:** $3,699. The specs assume RTX 5080 machines with 64 GB for most seats and one RTX 5090 machine for the character artist. They also include a Mac Studio M5 Max for iPhone and Vision Pro builds, an iPhone 18 Pro, a Rokoko capture suit bundle, and rented studio time for the main monologue. **Price list and buying schedule.** Each machine is bought the month its user is hired. Venue hardware is bought in months 7–8, once the venue agreement is signed. Every price is marked as published, estimated or needing a vendor quote. |Item|Year one| |---|---| |Core hardware and capture|$75k–$117k| |Software and services|$14k–$31k| |**Total**|**$88k–$148k**| |Contingency on top (10–15%)|$9k–$22k| |Optional headsets (Vive, Vision Pro, Galaxy XR)|$7.4k more, only if a venue or funder asks| **Year One Budget draft.** Hardware and software are filled in by quarter. Staff, contractors, legal, licensing, venue, travel and marketing are placeholder rows for the next step. Hardware and software alone nearly use up the $80k–150k "shared studio setup" figure in the pitch, so that number will likely need to rise once people costs are added. Three prices are placeholders until we get quotes or check them: - **Proto Luma hologram cabinet:** Proto only lists a $5,500 extended warranty publicly, so the cabinet itself will likely cost well above a flat screen. - **ArborXR headset management:** per-device pricing isn't public. - **ZBrush:** my estimate is unconfirmed. To start on people costs, I'd need a location (on-island, Boston or remote) and a salary or day-rate range for each role in the hiring plan. Sources: - [RTX 5090 price climb, Sept 2026 (AI.rs)](https://ai.rs/ai-for-business/rtx-5090-price-climb-september-2026) - [GPU memory crisis (Tech Times)](https://www.techtimes.com/articles/320169/20260711/gpu-memory-crisis-prices-rtx-5090-above-4300-nvidia-offers-paper-cards.htm) - [GPU relief pushed to 2028 (Tech Insider)](https://tech-insider.org/gpu-relief-pushed-2028-dram-price-jump-2026/) - [RTX 5080 price tracker](https://videocardprices.com/card/nvidia-rtx-5080/) - [RTX PRO 6000 price (Wccftech)](https://wccftech.com/nvidia-96-gb-rtx-pro-6000-blackwell-price-hits-13250-over-50-percent-hike/) - [DDR5 price index (Tom's Hardware)](https://www.tomshardware.com/pc-components/ram/ram-price-index-2026-lowest-price-on-ddr5-and-ddr4-memory-of-all-capacities) - [Puget Systems Unreal workstations 2026](https://www.pugetsystems.com/labs/articles/best-workstations-for-unreal-engine-in-2026/) - [Mac Studio M5 (9to5Mac)](https://9to5mac.com/2026/08/25/apple-unveils-next-generation-mac-studio-with-m5-max-and-m5-ultra/) - [iPhone 18 Pro (MacRumors)](https://www.macrumors.com/roundup/iphone-18-pro/) - [Pixel 11 (Google)](https://blog.google/products-and-platforms/devices/pixel/google-pixel-11-pro-xl/) - [Quest price increase (UploadVR)](https://www.uploadvr.com/meta-quest-3-quest-3s-price-hike-memory-crisis/) - [Vision Pro $3,699 (VR.org)](https://vr.org/articles/vision-pro-price-increase-3699-enterprise-signal-2026) - [Vive business store](https://shop-us.vive.com/collections/vive-focus-series-for-business) - [Galaxy XR (UploadVR)](https://www.uploadvr.com/samsung-galaxy-xr-google-android-xr-out-now/) - [Rokoko Full Performance Capture](https://www.rokoko.com/products/full-performance-capture) - [Unreal license](https://www.unrealengine.com/license) - [Unity 2026 pricing (80.lv)](https://80.lv/articles/unity-announces-its-upcoming-2026-price-changes) - [Marvelous Designer pricing (CG Channel)](https://www.cgchannel.com/2026/08/clo-virtual-fashion-releases-marvelous-designer-2026-1/) - [Maya 2027 pricing (CG Channel)](https://www.cgchannel.com/2026/03/autodesk-releases-maya-2027-and-maya-creative-2027/) - [Proto Luma store](https://shop.protohologram.com/collections/luma) - [ArborXR enterprise listing (CDW)](https://www.cdw.com/product/arborxr-device-management-subscription-1-year-enterprise-plan/7882197) --- > Spec a Looking Glass auto steroscopic display. a small one for development and the largest one for full-size figures (is it large enough)? --- Yes, the largest one is big enough for a full-size figure, but only just, and it isn't a true autostereoscopic display. I've added a new note, **Hardware — Looking Glass Displays**, with the full specs. **Looking Glass now sells two kinds of display:** - **Light Field Displays** are true autostereoscopic 3D. Each eye sees a different view of the rendered character, and Unreal has a plugin for them. The largest is only 27 inches ($10,000, sold out). - **Hololuminescent Displays (HLD)** come in 16, 27 and 86 inches. They show an ordinary 4K picture, including live Unreal output, in front of 3D depth built into the glass. That gives about two feet of perceived depth for a crowd with no glasses. But the character itself has no real parallax: you won't see her face from a new angle as you walk past. So only the 86-inch model can show a life-size person, and it's the HLD kind. **Small development unit:** - **Buy the 16-inch HLD ($2,000)** if the kiosk will use the 86-inch. It runs the same content pipeline, so what you author on it carries straight over. - If you want true light-field 3D for busts or objects like the telescope, the alternatives are the 16-inch Light Field Display ($3,000 on sale) or the 6-inch Looking Glass Go ($199). Both are currently sold out. **Large display: the 86-inch HLD ($20,000 list).** It's 4K, mounts in portrait, and is under 2.5 inches thick, so it hangs like a normal screen. It has a built-in camera, touchscreen, echo-cancelling microphone and speakers, which covers most of a conversational kiosk. **Is it big enough?** The picture area is about 75 × 42 inches (190 × 107 cm). How much room is left depends on the figure's height: |Figure height|Headroom left| |---|---| |5 ft 4 in|about 11 in| |5 ft 10 in|about 5 in| |6 ft|about 3 in, and a raised arm gets cropped| For life-size, the bottom edge has to sit at floor level. At 42 inches wide, a prop like the telescope or a second character won't fit at full scale. For a single Maria Mitchell kiosk it's a good fit. It's the only one of our three kiosk options that hangs flat on a historic wall. The other two are a flat 75-inch commercial screen ($2,000–3,200) and the Proto Luma cabinet (quote only). **What I updated:** - Both Looking Glass units are in the price list as optional items, which raises the optional total to $29,447. The core budget is unchanged. - The deployment specs, the Production Index and the dated sources page now include the Looking Glass options. Before buying, book a demo with our own Unreal character. Also ask Looking Glass for the recommended mounting height, brightness, and current lead time. Sources: - [Looking Glass HLD overview](https://lookingglassfactory.com/hld-overview) - [86" HLD product page](https://checkout.lookingglassfactory.com/products/hld-86) - [86" HLD announcement](https://blog.lookingglassfactory.com/introducing-the-86-hololuminescent-display-pre-orders-open-today/) - [86" HLD page](https://lookingglassfactory.com/86-hld) - [HLD docs](https://hlddocs.lookingglassfactory.com/) - [16" Light Field Display](https://checkout.lookingglassfactory.com/products/16-lightfield-display) - [27" Light Field Display](https://checkout.lookingglassfactory.com/products/27-lightfield-display) - [Looking Glass Go](https://checkout.lookingglassfactory.com/products/looking-glass-go) - [HLD vs light field (Tom's Hardware)](https://www.tomshardware.com/monitors/looking-glass-demos-hololuminescent-display-monitors-sizes-range-from-16-to-85-inches-starting-at-usd1-500) --- > Can we start by expanding on the skiilsets for Stage 2: likeness and body. Are there tutorials on Mesh to Metahuman, MetaHuman Creator, groom tools, Marvelous Designer to MetaHuman, and ZBrush to Maya (MetaHuman for Maya). What are the specific considerations for this pipeline, like what do you need to do in ZBrush to make a successful metahuman? These can be additional .md files in the appropriate sub-directories. --- Now the tutorial notes, using only links I found and checked.I've written up the Stage 2 skills and found tutorials for all five areas. They're 12 new or rewritten notes in your folder and the Claude project, and all links work. **Stage 2 skills note.** It sets out the order the tools are used in and seven skill areas. For each skill it marks what a new hire must already know versus what they can learn on the job. It includes four practical interview tests and says which roles hold which skills. **What to do in ZBrush.** ZBrush is used at two points with opposite rules, and the ZBrush note covers both. - **Before the character exists (the likeness sculpt):** - Work at real-world scale in centimetres, facing straight forward. - Keep a relaxed, symmetrical neutral face with lips fully sealed. - Eyes must be open with white sclera, and eyelids must not intersect the eyeballs. - Don't sculpt hair volume; hair, brows and beards become grooms later. - Keep the mesh clean with no holes. Export as OBJ if it's over about 200,000 vertices. - Give it a flat, unlit colour texture, and keep pores and wrinkles in separate texture maps. - For a full-body sculpt, use an A-pose, unclothed or in tight clothing. Loose skirts make the tool fit the garment instead of the body. - **After the MetaHuman exists (refining it):** - Never change the mesh's topology or vertex order; sculpt on layers. - Keep landmark vertices such as eyelid edges and lip corners on their features. - **Historical figures:** early daguerreotypes are often mirror-reversed, and the sculpt should match the program's date (Mitchell was 29 in 1847). **Considerations for each tool:** - **Mesh to MetaHuman:** the two conversion routes (head-only, or head and body, which is new in 5.8), their steps, and a table of what makes them fail. Because none of our figures can be scanned, the tool smooths the face toward real heads in its database. Distinctive features then have to be put back in Maya. - **MetaHuman Creator:** set the body first and the face second. Preview in the actual venue lighting, and decide early between the full-quality kiosk version and the lighter headset and phone version. - **MetaHuman for Maya:** its four tools and the route from ZBrush through Maya and back to Unreal. For facial expressions, the default for us is to let the software estimate them from the neutral face alone. The alternative, calibrating with about 12 of an actor's expressions, blends two identities and needs the actor's contract to cover it. - **Grooms:** strand hair for the kiosk and generated hair cards for headsets. It covers Maya, Blender and Houdini routes, the data Unreal requires from each groom, and period-accurate styling (under-groomed, not modern). - **Marvelous Designer:** version 2025.2 or later works with Unreal 5.6 and up. It covers the export route into a clothing asset that resizes to fit different bodies, plus period layering and fabric weights. **Tutorials.** There's a tutorial note for each area with only links I found this session: about 40 in total, a mix of Epic's documentation, community videos and a few paid courses. Each note ends with a suggested learning order and gaps I haven't filled yet, such as portrait-sculpting courses, Houdini grooms and period dress patterns. I have one correction to the budget: I costed Marvelous Designer at the $2,000 Enterprise licence. Studio use will likely need it, but the Personal licence terms should be checked before buying. Sources: - [Epic: From Mesh](https://dev.epicgames.com/documentation/en-us/metahuman/from-mesh) - [Epic: Neutral Pose](https://dev.epicgames.com/documentation/metahuman/neutral-pose?lang=en-US) - [Epic: From Custom Mesh](https://dev.epicgames.com/documentation/metahuman/metahuman-creator-from-custom-mesh-tool-in-unreal-engine?lang=en-US) - [Epic: MetaHuman for Maya](https://dev.epicgames.com/documentation/en-us/metahuman/metahuman-for-maya) - [Epic: Expression Editor](https://dev.epicgames.com/documentation/en-us/metahuman/expression-editor) - [Epic: Initial Head DNA](https://dev.epicgames.com/documentation/metahuman/obtaining-an-initial-metahuman-head-dna) - [Epic: Expressions for Likeness Calibration](https://dev.epicgames.com/documentation/metahuman/expressions-for-likeness-calibration?lang=en-US) - [Epic: Groom Starter Kit](https://dev.epicgames.com/documentation/metahuman/groom-starter-kit) - [Epic: Hair Card Generator](https://dev.epicgames.com/documentation/unreal-engine/hair-card-generator-for-grooms-in-unreal-engine) - [Marvelous Designer to MetaHuman USD workflow](https://support.marvelousdesigner.com/hc/en-us/articles/52699135975705-Marvelous-Designer-to-MetaHuman-USD-Garment-Integration-Workflow) - [Tailoring for MetaHumans demo (Epic)](https://www.unrealengine.com/news/tailoring-for-metahumans-clo-and-marvelous-designer-to-unreal-engine-demo) - [UE 5.8 Any Mesh to MetaHuman (YouTube)](https://www.youtube.com/watch?v=ZmiTuYglaRI) - [MetaHuman to ZBrush without breaking the rig (YouTube)](https://www.youtube.com/watch?v=IkLcTS4KFNo) - [Stylize MetaHumans with Maya (80.lv)](https://80.lv/articles/learn-how-to-customize-your-characters-with-metahuman-for-maya) - [MetaHuman parametric grooms with Blender (YouTube)](https://www.youtube.com/watch?v=ttoj9EHAn40) - [Custom clothes via LiveSync (Epic community)](https://dev.epicgames.com/community/learning/tutorials/V1m5/fab-create-custom-clothes-for-metahumans-5-6-via-livesync-marvelous-designer-and-unreal-engine-5-6) --- > in Maya/ can you make a list of ngSkinTools tutorials? Maybe create our own tutorial on a separate page for this skinning process and link back to the Maya tutorials page as a local tutorial. --- Now linking both pages from the Maya tutorials page and the index.I've added two pages in `02 Tools/maya/`: a list of ngSkinTools tutorials and our own skinning tutorial. **Maya — Tutorials** now links to both, with ours listed under a new "Local tutorials" section. Both pages are also in your folder, the project and the Production Index. In Maya 2027, ngSkinTools is now built in under the name **Skin Tools**. Maya 2026 still uses the ngSkinTools 2 plugin. The tutorial list covers both. **ngSkinTools — Tutorials:** - **From Autodesk:** the Maya 2027 Skin Tools help pages and a skinning demo video. - **From ngSkinTools:** five short official videos (Layers Basics, Mirroring, Smoothing, Interactive Joint Selection, and a full character body), plus the user guide, FAQ and plugin download. - **Community:** Rigmarole Studio's skinning tips and a few more videos. - A suggested learning order. **Tutorial — Skinning Garments and Props to a MetaHuman** walks through attaching period garments and props to the MetaHuman skeleton: - **When to use it:** spectacles, caps, collars, cuffs, and props with joints such as a telescope stand. Never use it on the MetaHuman body or face; the facial rig depends on their existing weights. Garments that can go through Marvelous Designer should go that way instead. - **The steps:** scene prep, bind settings, building layers, testing through a set of poses (including sitting and a bow), cleanup for Unreal, and import onto the MetaHuman skeleton. - A checklist and a troubleshooting table. The tutorial is marked as a draft. I wrote it from the documentation, not from running it in Maya. Test it on the first real garment and correct any menu names that differ. Sources: - [ngSkinTools v2 tutorials](https://www.ngskintools.com/v2/tutorials/) - [ngSkinTools v2 user guide](https://www.ngskintools.com/documentation/v2/userguide/) - [ngSkinTools v2 What's New](https://www.ngskintools.com/documentation/v2/userguide/version2-whats-new/) - [Maya 2027 Skin Tools (Autodesk)](https://help.autodesk.com/cloudhelp/2027/ENU/Maya-WhatsNew/files/GUID-46722737-9D1C-41BA-85D0-83F7F3D13682.htm) - [Skinning Workflow Demo, Skin Tools in Maya 2027 (YouTube)](https://www.youtube.com/watch?v=XOmRKZnyIqQ) - [ngSkinTools Skinning Tips (Rigmarole Studio)](https://rigmarolestudio.com/ngskintools-skinning-tips/) - [Skinning a full character with ngSkinTools (YouTube)](https://www.youtube.com/watch?v=Xu1qXBgU6sQ) --- > Let's do a personnel budget estimate. What are the going rates for a remote position for each of the roles we have specified? The current Augmented Traveler team is two founders, CEO and CTO, who do not live in the same city. What would the roll-out schedule be for each position, it's skills and requirements (I believe you have specified these somewhat), detailed in separate .md documents? provide a budget summary for personnel, then fold that into a combined summary (do not add personnel to the hardware/software docs as they exist.) --- Your folder is updated. Now the Claude project.Personnel for year one comes to about **$621,000–$1.15 million**, not counting founder pay. Added to hardware and software, the combined total is about **$718,000–$1.33 million**. I left the hardware and software budget note unchanged and put the combined figures in a separate summary note. I assumed the CEO covers the executive-producer role and the CTO covers the technical-director role, so neither is a hire. Founder pay isn't included. Rates are US remote averages, and employees carry an extra 25–35% for payroll taxes and benefits. **Core hires (remote employees):** |Role|Starts|Base salary (full-time)|Year-one cost| |---|---|---|---| |AI / conversation engineer|Month 1|$110k–$160k|$126k–$198k| |Conversation designer (half-time)|Month 1|$70k–$100k|$40k–$62k| |Character artist|Month 2|$90k–$130k|$94k–$146k| |Technical animator (three-quarter time)|Month 2|$85k–$125k|$66k–$105k| |Unreal engineer|Month 3|$106k–$157k|$99k–$159k| **Everyone else:** - **Contractors, paid hourly:** historian, corpus editor, groom artist, costume designer, cloth artist, environment artist, XR engineer, mobile AR engineer and DevOps. - **Performance:** voice director, a voice actor and a motion-capture actor at SAG-AFTRA union rates, a mocap technician and a sound designer. - **Deployment:** kiosk installer, accessibility designer and QA tester. - **Optional:** a Vision Pro developer. Each role, founders included, now has its own note in `roles/positions/` (26 in all). Each gives start month, type of engagement, rate and where it came from, year-one cost, responsibilities, skills, requirements and how the role works remotely. The personnel budget note has the full hiring timeline, a Gantt chart and monthly headcount: it peaks at 14 people in month 4 and again in month 7. **Combined year-one budget:** |Line|Low|High| |---|---|---| |Hardware and capture|$74,792|$117,143| |Software and services|$13,654|$31,054| |Contingency on hardware and software|$8,845|$22,230| |Personnel, not counting founders|$620,799|$1,154,986| |**Total**|**$718,090**|**$1,325,413**| People are about 86–87% of the cost. Recruiting, a payroll service and two in-person team weeks are included in personnel. **Still not included:** founder pay, rights counsel, collections licensing, venue costs and marketing. **This is well above the pitch.** The pitch shows about $230k–$400k (studio setup plus the Mitchell production). That figure assumed part-time help; this estimate uses full remote market rates and five employees. You'll need to choose which scenario the pitch shows before it goes to the head-hunter. The personnel note lists ways to cut the total. The biggest are contracting the artist and animator for the first production and using non-union performers. Sources: - [Glassdoor remote LLM engineer](https://www.glassdoor.com/Salaries/remote-llm-engineer-salary-SRCH_IL.0,6_IS12228_KO7,19.htm) - [Glassdoor conversation designer](https://www.glassdoor.com/Salaries/conversation-designer-salary-SRCH_KO0,21.htm) - [Glassdoor 3D character artist](https://www.glassdoor.com/Salaries/3d-character-artist-salary-SRCH_KO0,19.htm) - [Glassdoor technical animator](https://www.glassdoor.com/Salaries/technical-animator-salary-SRCH_KO0,18.htm) - [ZipRecruiter remote Unreal developer](https://www.ziprecruiter.com/Salaries/Work-From-Home-Unreal-Engine-Developer-Salary) - [ZipRecruiter consulting historian](https://www.ziprecruiter.com/Salaries/Consulting-Historian-Salary) - [Upwork 3D artist rates](https://www.upwork.com/hire/3d-artists/cost/) - [Glassdoor environment artist](https://www.glassdoor.com/Salaries/environment-artist-salary-SRCH_KO0,18.htm) - [Glassdoor XR developer](https://www.glassdoor.com/Salaries/xr-developer-salary-SRCH_KO0,12.htm) - [Glassdoor iOS developer](https://www.glassdoor.com/Salaries/ios-developer-salary-SRCH_KO0,13.htm) - [Glassdoor DevOps engineer](https://www.glassdoor.com/Salaries/devops-engineer-salary-SRCH_KO0,15.htm) - [Glassdoor game tester](https://www.glassdoor.com/Salaries/game-tester-salary-SRCH_KO0,11.htm) - [SAG-AFTRA Interactive Media wage table](https://www.sagaftra.org/sites/default/files/2025-2028%20Interactive%20Media%20Agreement%20Wage%20Table.pdf) - [YunoJuno sound designer rates](https://www.yunojuno.com/freelancer-rates-job-role/sound-designer) - [BLS Employer Costs for Employee Compensation](https://www.bls.gov/news.release/ecec.nr0.htm) --- > thanks for keeping the citations up-to-date. Can we go with two numbers for the personnel budget? One with full-time remote employees and another with contract labor for the amount of time necessary to perform the task: modeling, marvelous designer, writers, motion capture, actors, and others. I imagine the Unreal, AI engineers, and dev-ops would be full-time, which is still almost 5 employees! hm. --- Good idea. I'll split the model into two scenarios and regenerate the role notes and both budget summaries.Your folder first, then the project.I've split the personnel budget into two scenarios. In the lean one you have exactly five full-time people: the two founders plus the Unreal, AI and DevOps engineers. |Year one (excluding founder pay)|A: full-time studio|B: lean core + contracts| |---|---|---| |Full-time people, including founders|13|5| |Employees|$879k–$1.37M|$317k–$492k| |Contracts and fees|$69k–$186k|$237k–$583k| |Recruiting, payroll service, team travel|$31k–$65k|$14k–$30k| |**Personnel total**|**$978k–$1.62M**|**$568k–$1.10M**| |**Combined with hardware and software**|**$1.08M–$1.79M**|**$665k–$1.28M**| **Scenario A** makes 11 roles full-time employees. That includes one new role, a character technical artist who handles both hair and clothing. Research, costume, performers and installation stay on contract. **Scenario B** contracts everyone else for only the hours their task needs. For example, the character artist gets 480–640 hours over months 2–7, and the writer 360–520 hours spread across the year. Scenario B saves about $410k–$514k. B's high end is still over $1M because I used wide hourly rates. The top of each contract rate range is well above what an employee in the same job earns per hour. Once you get real quotes, the range will narrow. **Two things to consider:** - **DevOps as a full-time hire** costs $92k–$135k for the eight months it's needed. If the CTO or AI engineer can cover infrastructure until the kiosk opens, a part-time contract ($13k–$29k) would save most of that. - **A middle path:** add the character artist as a fourth full-time hire. Likeness is the hardest skill to contract well, and every future figure needs it. **What changed in the files:** - Each of the 27 role notes now shows its engagement type, timing, rate and year-one cost under both scenarios. - The personnel budget note has a timeline chart for each scenario, a quarterly spend table, monthly headcount, and a trade-offs table covering speed on the second figure, keeping knowledge in-house, and management load. - The combined summary shows both totals. The pitch's $230k–$400k needs to rise to at least scenario B's low end of about $665k, unless the first production is scaled down. The hardware and software note is unchanged. No new sources were needed, so the citations page is unchanged too. --- > Do these full-time rates include typical American employee costs, like Federal Withholding, Social Security taxes? I am going to go with each person is their own LLC or Sole Proprietorship for the first year. What would the additional costs for a company to hire actual full-time employees? You don't have to generate any new .md files for this question. --- Yes. The employee figures already add 25–35% on top of salary to cover the employer's side. The Social Security figure I checked (a $184,500 cap for 2026) supports that range. Federal income-tax withholding isn't an extra cost to the company, though: it comes out of the employee's own salary, so it's already inside the quoted pay. **What a company pays on top of salary for a real W-2 employee** (example: $130,000 salary) |Cost|Rate|On $130k| |---|---|---| |Social Security (employer half)|6.2% of wages up to $184,500 in 2026|$8,060| |Medicare (employer half)|1.45%, no cap|$1,885| |Federal unemployment tax|about 0.6% of the first $7,000|$42| |State unemployment tax|varies by state; new employers often pay about 1–4% on the state's wage base|roughly $300–$2,000| |Workers' compensation|low for remote office work|roughly $150–$650| |**Required by law, subtotal**|**about 8–10%**|**about $10,500–$12,600**| |Health insurance (optional but expected)|KFF 2025 average single premium is $9,325; the employer usually pays most of it|roughly $7,500–$8,000 single; far more for family coverage| |401(k) match (optional)|commonly 3–4%|$3,900–$5,200| |Payroll service or PEO|about $50–$150 per employee per month|$600–$1,800| |**Typical total**|**about 20–35% of salary**|**about $23,000–$28,000**| Paid time off doesn't add cash. It's time you pay for without getting work, which matters for scheduling. **If each person is their own LLC or sole proprietor (1099 contractors)** - **What goes away:** the company pays none of the costs above and handles no withholding. You collect a W-9 from each person and file a 1099-NEC for anyone paid $600 or more. - **What they pay instead:** contractors pay both halves of Social Security and Medicare themselves (15.3%) and buy their own insurance. Experienced people usually charge 20–40% more than the employee salary for the same job to cover that. - **Effect on the budget:** in Scenario B, the Unreal, AI and DevOps engineers were priced as employees with the added costs. As contractors, their cost will likely come out close to the same, because the rate goes up by roughly what the load was. The other contract lines in both scenarios already assumed contractor rates. - **Contracts need:** a clear statement of work, assignment of intellectual property to the company (important for character assets and code), confidentiality, and the contractor's own insurance where relevant. **A risk to check before committing: misclassification.** Someone working full-time, ongoing, and directed by the company may legally be an employee even if they have an LLC. - **Massachusetts** uses a strict "ABC test." Among other things, the work must be _outside the usual course of the company's business_. Character artists and engineers at a character studio likely fail that part. - **Other states:** California and several others have similar tests. What matters is where each worker lives, not only where the company is. - **Penalties** can include back payroll taxes, overtime and fines. - **Who to ask:** an employment attorney or CPA before signing. I'm not a lawyer or accountant. - **Safer options:** - Keep the contractors genuinely project-based: defined deliverables, their own tools and schedule, and several clients. - Use an employer-of-record service or PEO for anyone effectively full-time. It adds the employer costs above, but it handles multi-state payroll and removes the classification risk. As you asked, I haven't changed any files. If you decide to go with LLC contractors, I can re-price the three engineers in Scenario B at contractor rates in the personnel budget. Sources: - [2026 Social Security wage base (Patriot Software)](https://www.patriotsoftware.com/blog/payroll/social-security-wage-base/) - [FICA tax rates 2026 (OnPay)](https://onpay.com/insights/what-are-fica-tax-rates/) - [KFF 2025 Employer Health Benefits Survey](https://www.kff.org/health-costs/2025-employer-health-benefits-survey/) - [KFF 2025 summary of findings (PDF)](https://files.kff.org/attachment/Employer-Health-Benefits-Survey-2025-Annual-Survey-Summary-of-Findings.pdf) - [BLS Employer Costs for Employee Compensation](https://www.bls.gov/news.release/ecec.nr0.htm) - [Massachusetts ABC test (Brandon J. Broderick)](https://www.brandonjbroderick.com/massachusetts/abc-test-and-ma-employee-classification-why-it-matters) - [Massachusetts independent contractor law (Steffans Legal)](https://www.steffanslegal.com/misclassified-independent-contractors-insights) --- You've hit your session limit · resets 8:20pm (America/Chicago) --- > Can you place all this information and sources in a self-employment .md file for now? --- Ran 5 commands, used 5 tools, created a file I've added **Self-Employment — Contractors vs Employees.md** in `01 Start Here/budget/`. It's saved in the project, in your folder on the E: drive and in outputs. It covers: - **What a full-time employee costs beyond salary.** For a $130,000 employee, the taxes required by law add about 8–10%. With benefits and payroll fees, the usual total is about 20–35%, or roughly $23,000–28,000. - **What changes when each person is their own LLC.** Each contractor pays the 15.3% self-employment tax and their own benefits, and typically charges 20–40% more than the equivalent salary to cover that. So for the three full-time engineers in Scenario B, the cost ends up about the same either way. - **What the studio still has to do.** Collect a W-9, file a 1099-NEC for anyone paid $600 or more, and sign a contract covering the work, IP ownership, confidentiality and insurance. - **The risk of calling employees contractors.** Massachusetts has a strict test for who counts as a contractor. It's especially hard to pass for full-time engineers doing the studio's core work. The note ends with lower-risk options and a reminder to check with an employment attorney or CPA. The seven sources are listed in the note and added as a new section of the Sources — 2026-09-19 page. The new note is also linked from the Production Index and from the Personnel budget's assumptions. The hardware/software budget draft is unchanged. --- > What are all of the ZBrush and Maya tutorials that we might need for this project? We have one on skinning. What about creating some local tutorials on HumanIK mocap cleanup, USD for Maya, or Obtaining Initial Head DNA, and Expression Editor fundamentals? For ZBrush, how to start modeling a head and base body, and "learn the no-topology-change round trip" as it relates to a MetaHuman workflow. be sure to include new tutorials in the local tutorials section of the -Tutorials .md file. --- I wrote all six local tutorials you asked for. Each is listed in the "Local tutorials" section of Maya — Tutorials or ZBrush — Tutorials. All the new and changed notes are saved in the project, in your folder on the E: drive and in outputs. The link check found no problems in anything new. **New Maya tutorials** (in `02 Tools/maya/`): - **Tutorial — Obtaining Initial Head DNA:** the route from the ZBrush sculpt, through Mesh to MetaHuman and Creator, to the head DNA loaded in Maya. It also covers file naming and why you keep the first export. - **Tutorial — Expression Editor Fundamentals:** how the editor works, the three editing modes, the order of work and how to export back to Creator or Unreal. It uses MH1, the likeness level that doesn't need a scan, as our default. It also flags an Unreal 5.8 trap: deleting the old DNA asset before re-importing breaks the rig with no error message. - **Tutorial — HumanIK Mocap Cleanup for MetaHuman:** retargeting a capture onto the MetaHuman skeleton, fixing feet, hips, hands and props on separate animation layers, and exporting to Unreal. It includes a joint-mapping table, and a warning not to retarget in Unreal and then re-apply the take in Maya. - **Tutorial — USD for Maya:** props with variants, such as a ledger open or closed. It covers export to Unreal, Vision Pro and iPhone. Characters and garments stay out of USD. **New ZBrush tutorials** (in `02 Tools/zbrush/`): - **Tutorial — Starting a Head and Base Body in ZBrush:** recommends starting from the nearest MetaHuman rather than a blank sphere. It covers reference photos, real-world scale, blocking in the head and a custom A-pose body. - **Tutorial — ZBrush No-Topology-Change Round Trip:** the list of ZBrush operations that break the rig, sculpting on layers, and a quick test in Maya that catches a broken mesh. It ends with bringing the sculpt back into the Expression Editor. **Tutorials we might need:** each Tutorials note now ends with a full table of what the project needs. Each row is marked as written, covered by an external tutorial, or still to write. Still to write: - Maya: Pose Editor correctives under period tailoring, a hand-pose library, FBX export presets and Python batch export. - ZBrush: sculpting expression targets, polypaint albedo export and baking detail maps. Portrait likeness from photographs and age changes still need research. All six tutorials are drafts written from the current Epic, Autodesk and Maxon documentation. Whoever first runs each one on a real figure should correct the menu names in it. I also linked the tutorials from the Production Index, Maya — Process and the two considerations notes. The research sources are added as a new section of Sources — 2026-09-19. One version correction: USD for Maya is now at 0.37 (July 2026), and the USD tutorial notes it. --- > please write the "tutorials we might need" for Maya and the ZBrush. register them with all the .md files. next, we can do the research needed on ZBrush. --- Now the four ZBrush tutorials.Now the 12 position notes.The eight remaining Maya and ZBrush tutorials are written and registered, and the ZBrush likeness and age research is done as a new note. Everything is saved in the project, your E: drive folder and outputs, and the link check passes. The new tutorials are drafts from the Epic, Autodesk and Maxon documentation; none has been run in the software yet. **New Maya tutorials:** - **Pose Editor Correctives for Period Tailoring** – fixes shoulders and elbows that break under period sleeves and bodices. It sticks to edits that can still go back into MetaHuman Creator. - **Hand Pose Library for Gestures** – period-appropriate hand poses, saved in Studio Library (a free Maya pose tool) and sent to Unreal as a Pose Asset. That asset serves both mocap cleanup and live gestures. - **FBX Export Presets for Unreal** – three shared presets: static props, skinned meshes and animation. - **Maya Python Batch Export and Naming Checks** – a script that checks names, units, history and transforms before exporting. It parses as valid Python but hasn't been run inside Maya. **New ZBrush tutorials:** - **Sculpting Expression Targets for the Expression Editor** – when a single expression should be sculpted, and when it shouldn't. - **Flat Albedo from Polypaint** – the unlit skin colour map. It covers the extra care needed on skin tone. - **Baking Detail Maps for MetaHuman** – normal and cavity maps loaded into Creator. - **ZBrush Python Export Presets** – optional. One interface call in it still needs checking against Maxon's documentation. **Where they're registered:** - **Tutorials notes:** both have the new "Local tutorials" rows, and every "to write" item is now marked done. - **Other notes:** the Production Index, the Stage 2 skills note, the process and considerations notes, and cross-links between the tutorials. - **Position notes:** 12 roles now have a "Training (local tutorials)" section listing the ones they should work through. I added this to the script that generates those notes, so a rebuild keeps it. I checked that the rebuild left the personnel budget unchanged. - **Sources page:** new dated sections for the tutorials and the research. **ZBrush research** is in the new note **ZBrush — Likeness and Age Research**: - **Reading early photographs:** most daguerreotypes are mirror images. The plates also recorded colour badly: warm skin came out dark, blue eyes pale, and freckles stronger than in life. So skin, eye and hair colour must come from written descriptions, not the photos. - **Evidence vs. interpretation:** the note sets which features can be taken from photos and which need a label saying they're interpreted. - **Age:** the note gives a layered order for making a face younger or older. Mitchell at 29 means working back from photos that start around age 33. - **Rights:** the Maria Mitchell Association grants image use once, for one format only. A 3D likeness shown on a kiosk, a headset and a phone needs a broader licence. Negotiate one before sculpting from their photographs, or work only from public-domain copies. This should go on the rights map. The likeness and age topics are covered by rules and paid courses for now. They can become step-by-step local tutorials once the first sculpt is under way. --- Claude writes .md files to the hard drive. Beyond Compare copies the new files to the Obsidian vault: ![[2026-09-19 21_34_32-Conversation - Claude - PubProjects - Obsidian 1.13.7.png]] Obsidian publishes the files to the website: ![[2026-09-19 21_40_28-Tutorial — Starting a Head and Base Body in ZBrush - PubProjects - Obsidian 1.13.png]] --- > lets do the Marvelous Designer local tutorials. > Can we create a Getting Started tutorial, with links to other tutorials and videos? > Then, we should tackle Unreal: getting started, sequencer, blueprints (isn't blueprints going away with the switch to UEFN?), OpenXR, packaging, Pixel Streaming, authoring a timed experience, authoring a Q&A experience with ambient living movement by the virtual human while waiting for a question. what all does the project need? how does Marvelous Designer interact with Unreal? Or is that handled by MetaHuman? How does MetaHuman implement in Unreal? --- Now the 19 position notes.I wrote 5 Marvelous Designer tutorials and 9 Unreal tutorials. Each tool's Tutorials note now has a local-tutorials table and a full list of what the project needs. Everything is in the project, your E: drive folder and outputs, and every link resolves. The tutorials are drafts from vendor documentation. Menu names should be checked the first time someone runs each one. **Marvelous Designer** (`02 Tools/marvelousDesigner/`): - **Getting Started with Marvelous Designer** – the hub. It covers what MD does for us and how to set up a seat. It lists MD's official Quick Start videos in order, plus paid courses. A one-hour practice skirt teaches pattern, sew, arrange, simulate and fabric. It ends with the path through our other tutorials. - **MetaHuman Avatar into Marvelous Designer** – brings the figure's own body into MD (MD 2025.2+ can read MetaHuman files directly). - **Drafting a Period Garment from a Historical Pattern** – pattern sources (Janet Arnold, Norah Waugh, reproduction patterns, Nantucket Historical Association garments), then tracing, sizing to the figure, sewing and fabrics. - **Period Layers, Hoops and Skirts** – simulate the outer garment and fake the layers underneath. Epic has no documented fix for hoop skirts, so the plan is to test on the first skirt. Headsets and phones start with non-simulated clothing. - **Marvelous Designer Garment to MetaHuman Outfit** – the full chain from MD into MetaHuman Creator. **Licence:** MD's cheaper Personal licence can't be used by a studio; MD can revoke it. The budget already has the Enterprise licence at $2,000/yr. A new Indie licence may be cheaper, so ask for a quote. **Unreal** (`02 Tools/unreal/`): Getting Started, MetaHumans in Unreal, Blueprints, Sequencer for Monologues, Authoring a Timed Experience, Authoring a Q&A Experience with Ambient Idle, OpenXR Headset Builds, Packaging Builds for Each Surface, and Pixel Streaming 2. **Your questions:** - **Is Blueprint going away?** Not for us. Epic plans no UE5 release after 5.8, and UE6 (UE5 plus Fortnite's editor, UEFN) is targeted for early access at the end of 2027. Epic says Blueprints will be in early UE6 and phased out later, with conversion tools. UEFN uses Epic's newer Verse language instead of Blueprints, which is where the idea comes from. We stay on 5.8 with Blueprints and C++. We keep scheduling logic in data-driven systems (Epic's State Tree and data assets), which should convert more easily later. - **Marvelous Designer, MetaHuman and Unreal:** MD makes the clothes and exports them. Unreal turns the garment into cloth that can simulate and resize to a body, called an Outfit Asset. MetaHuman Creator puts that outfit on the figure. MD never talks to MetaHuman directly. Resizing happens in the editor and is baked into the build, not done live. - **How a MetaHuman works in Unreal:** you edit the character in MetaHuman Creator inside Unreal, then "assemble" it into a ready-to-run character. We need two versions: full quality for the kiosk (about 1–2 GB) and a light one under 100 MB for headsets and phones. The assembled character bundles face, body, hair, clothing, facial rig and live-animation input. We wrap it rather than edit it, so re-assembling doesn't wipe our additions. - **What the project needs:** the full list is in the Tutorials notes. Still to write for Unreal: - source control setup; - the conversation client plug-in; - captions UI; - kiosk lighting; - place kits; - Vision Pro; - performance profiling. **Risks before building on these:** - **Headset plug-ins:** Meta's plug-in labelled for 5.8 has an open bug, and Google's Android XR guidance still targets UE 5.6. Start headsets on Epic's built-in OpenXR plug-in. - **Live lip sync:** Epic doesn't say whether its real-time audio-driven face animation works in a packaged build. Test it and NVIDIA's Audio2Face on the kiosk PC before choosing. I also added training lists to 19 position notes (through the generator, and the personnel budget came out unchanged). The Unreal process note now uses State Tree for the program clock and records the UE6 plan. The new sources are on the dated Sources page. --- > Please create those missing Unreal tutorials: > > - source control setup; > - the conversation client plug-in; > - captions UI; > - kiosk lighting; > - place kits; > - Vision Pro; > - performance profiling. > > then, generate a list of tutorials needs for Groom, MoCap, and AI mind (I have some questions about this process and tools like SillyTavern and using open weight models for local inference). --- The seven missing Unreal tutorials are written, and the Groom, Mocap and AI Mind notes now each end with a full list of tutorials the project needs. Everything is saved in the project, your folder on the E: drive and outputs, and all the links resolve. The tutorials are drafts from vendor documentation; none has been run on real hardware yet, and the plug-in code hasn't been compiled. **New Unreal tutorials** (`02 Tools/unreal/`): - **Source Control Setup** – recommends Perforce P4 (formerly Helix Core), which is free for up to 5 users and 20 workspaces if you host it. That covers the founders and the three full-time engineers. Epic's new open-source tool, Lore, isn't ready: Epic's own roadmap says it isn't production-ready and puts its editor plug-in in 2027. - **Conversation Client Plugin** – the message format to agree with the AI engineer, plus a C++ skeleton for microphone capture, streaming voice playback, timeouts and offline fallback. Check the code against Epic's API pages on the first build. - **Captions UI** – caption standards from WCAG, the FCC, DCMP, the Smithsonian and Meta, a style guide for each surface, captions for scripted monologues, and a live caption display for answers. - **Kiosk Lighting for a Life-Size MetaHuman** – period lamp and daylight, making sure the eyes show a highlight, and calibrating the display. No published benchmarks exist for this GPU class, so it includes a benchmarking step. - **Place Kits for Historic Interiors** – rooms built from Library of Congress measured drawings (HABS, which includes the Swain-Mitchell house) as reusable modular pieces, with pre-baked lighting for headsets. - **Vision Pro Builds** – Epic said in January 2026 that it has no roadmap to take Vision Pro beyond experimental, and Unreal can't do mixed reality there. The tutorial recommends a native Apple (RealityKit) or Unity route if a partner ever wants the figure standing in a visitor's own room. - **Performance Profiling** – frame budgets (Quest at least 72 fps), Unreal's and Meta's profiling tools, and where MetaHumans cost the most. It has a table to record real measurements. **Tutorials still needed (not written yet):** - **Grooms:** - period hairstyles from evidence; - women's styles in Maya XGen; - period facial hair (Epic has no doc on facial hair); - hair under caps and bonnets; - lighter hair versions for headsets. - **Mocap:** - running a capture day; - Epic's new experimental single-camera body capture; - a body gesture library with tags; - idle takes; - period movement coaching. Open question for counsel: which SAG-AFTRA contract covers a museum kiosk, since it isn't a video game. - **AI Mind:** about a dozen, covering the corpus, retrieval, persona, the policy layer, evaluation, speech in and out, and privacy. Three are waiting on your decisions: cloud or local model, local inference, and SillyTavern. **For your local-inference and SillyTavern questions**, a starting point is in AI Mind — Tutorials: - **Models:** open-weight families that fit a single RTX 5090 (32 GB), such as Gemma 4 and Qwen, are Apache-licensed. A 20–35B model compressed to 4–5 bits leaves room for speech, voice and lip sync on the same card. The newest model names and the speed figures come from single recent sources; benchmark them on our hardware. - **SillyTavern** is a good sandbox for trying out a persona. I'd keep it out of the kiosk: its own docs warn against exposing it publicly, it can't do the historian-controlled sourcing with citations and refusals that the Ground Rules require, and its AGPL licence would oblige us to publish our changes. Ask your questions whenever you're ready, and I'll turn the answers into the decision note and the local-inference tutorial. --- > get on these tutorials: > > - Grooms: > - period hairstyles from evidence; > - women's styles in Maya XGen; > - period facial hair (Epic has no doc on facial hair); > - hair under caps and bonnets; > - lighter hair versions for headsets. > - Mocap: > - running a capture day; > - Epic's new experimental single-camera body capture; > - a body gesture library with tags; > - idle takes; > - period movement coaching. > > AI Mind: about a dozen, covering the corpus, retrieval, persona, the policy layer, evaluation, speech in and out, and privacy. Three are waiting on your decisions: cloud or local model, local inference, and SillyTavern. anything from the "4. Every AI-mind tutorial the project needs" for which you can find reasonable research. ![[2026-09-20 09_33_43-Conversation - Claude - PubProjects - Obsidian 1.13.7.png]] ![[2026-09-20 09_32_31-NVIDIA GeForce Overlay.png]] > be sure that you are writing a citations .md for today's tasks! > --- Claude asks permission to access a website: ![[2026-09-20 09_46_06-NVIDIA GeForce Overlay.png]] Claude spins up multiple agents: ![[2026-09-20 09_33_43-Conversation - Claude - PubProjects - Obsidian 1.13.7 1.png]] Claude keeps track of progress on the project: ![[2026-09-20 09_37_08-NVIDIA GeForce Overlay.png]] Claude is working: ![[2026-09-20 09_44_51-NVIDIA GeForce Overlay.png]] I am almost out of credit for this session: ![[2026-09-20 09_44_58-NVIDIA GeForce Overlay.png]] and, that's it: ![[2026-09-20 09_48_50-NVIDIA GeForce Overlay.png]] --- Now sync all changed files.Now the 22 position notes.I've written all 24 tutorials: 5 on grooms, 5 on mocap and 14 on the AI mind. Every item on the three "tutorials the project needs" lists is now marked written. They're saved in the project, your E: drive folder and outputs, and all links resolve. On your reminder about citations: today's page is **Sources — 2026-09-20**. It now has separate sections for grooms, mocap and the AI mind, about 190 source links in all, alongside this morning's Unreal sources. The citations index lists everything covered today. These are drafts from vendor and archive documentation. None has been tested in the software yet, and the Python snippets pass a syntax check but haven't been run. **Grooms:** period hairstyles from evidence, XGen period hair for MetaHuman, period facial hair, hair under caps and bonnets, and hair cards for headsets. - Epic's Groom Starter Kit uses XGen's older Description system, not Interactive Groom, and the exporter only supports the older system. The XGen tutorial follows that. - Epic has no facial-hair guide, but MetaHuman Creator does have separate beard and moustache slots. The facial-hair tutorial is built on those. **Mocap:** running a capture day, single-camera body capture, a gesture library with tags, capturing idle takes, and period movement coaching. Some advice is our own and is labelled that way, including suit calibration, clip lengths and the pass marks for comparing the single-camera capture with a suit. Epic's single-camera capture is still experimental, and several of its details come only from forums. **AI Mind:** 14 tutorials, starting from **How the Mind Works** as the overview. The three topics that were waiting on decisions are written as decision aids: - **Cloud or local models:** it sets out four options and a way to estimate cost. It suggests a cloud model by default, with a local model when offline and for school visits. If our benchmark shows a local model scoring close to the cloud one, it suggests going fully local instead. It ends with seven questions only you and the CTO can settle. - **Local inference:** Ollama or llama.cpp on the kiosk, how to fit the model alongside Unreal and speech on one graphics card, and which benchmarks to run. - **SillyTavern:** a sandbox for the conversation designer and historian to try out a persona. It is never public-facing. **Findings that change plans:** - **Actor voice:** ElevenLabs won't let us clone someone else's voice at the high-quality tier, even with their consent. The actor has to create it on their own account and share it with us, so the voice contract must cover that. - **Gemini:** Google's API terms bar services likely to be used by under-18s, which likely rules it out for kiosks. Confirm with counsel. - **Children's voices:** the amended COPPA rule treats voice recordings as personal information. Our push-to-talk, keep-no-audio design fits the narrow exception for answering a child's request. Whether COPPA applies to a kiosk at all is for counsel. - **Mitchell's plain speech:** I found no example of Mitchell herself using "thee" or "thy". The historian needs to settle how the persona uses plain speech. - **Absalom Boston:** no writing in his own hand turned up. He should stay a retrieval-only figure with pre-written bridge answers unless the Nantucket Historical Association finds some. Two smaller updates: 22 position notes now list their new training, and the personnel budget is unchanged. Ask your local-model and SillyTavern questions whenever you're ready, and I'll fold the answers into the decision note. ---