Skip to content

Open-source platform for 3D AI agents. Turn text or a photo into a rigged, animated GLB avatar, give it an LLM brain, memory and a wallet, and embed it anywhere with one web component. Remote MCP server, x402 payments, Three.js, browser-native.

docs(nvidia): turn the Inception catalog request into a send-ready packet

nirholas2ab05488b370
diff --git a/docs/nvidia-apps-catalog-listing.md b/docs/nvidia-apps-catalog-listing.md
index 999dbcebb..62d3012f0 100644
--- a/docs/nvidia-apps-catalog-listing.md
+++ b/docs/nvidia-apps-catalog-listing.md
@@ -8,7 +8,7 @@ Related docs: [nvidia-inception.md](./nvidia-inception.md) (membership overview)
## Status: product record filed, public listing not live (2026-08-14)
-The portal shows the product row (`three.ws | Shipping | GPU Accelerated`), but a search for `three.ws` on the public catalog returns 0 results. That is expected, not a bug: **the portal record and the public catalog are two different systems.**
+The portal shows the product row (`three.ws | Shipping | GPU Accelerated`), but a search for `three.ws` on the public catalog returns 0 results. **Re-verified 2026-09-04:** still 0 results on the public catalog, and the portal record is still the uncorrected one below (Used: cuBLAS, CUDA Python, CUDA Toolkit, cuDNN; Considering: DeepVariant NIM, Riva, TensorRT, Triton, Omniverse Kit, Audio2Face), so neither step 1 nor step 2 has happened yet. That is expected, not a bug: **the portal record and the public catalog are two different systems.**
- **Portal > Profile > Products** makes the company benefits-eligible and feeds NVIDIA's internal recommendation engine. The portal's own wording is "we **may** feature it in the personalized recommendations we share with our customers."
- **The public catalog** ([marketplace.nvidia.com/en-us/enterprise/applications](https://marketplace.nvidia.com/en-us/enterprise/applications/), filtered view for startups at [nvidia.com/en-gb/accelerated-applications/inception](https://www.nvidia.com/en-gb/accelerated-applications/inception/)) is curated and published by NVIDIA. Submissions are reviewed on a rolling basis and approval is subject to availability and fit. Nothing in the portal auto-publishes.
@@ -83,7 +83,7 @@ Also do not claim the **video-to-scene world scanner** or the **texture worker**
- **CUDA** with custom-compiled extensions (nvdiffrast, diffoctreerast, diff-gaussian-rasterization, torchmcubes)
- **NVIDIA Kaolin** (TRELLIS worker)
- **NIM APIs** (integrate.api.nvidia.com): Nemotron Super and Nano LLMs, Nemotron Nano VL vision, NV-EmbedQA-E5-v5 embeddings, Mistral 4B reranking, NemoGuard 8B content safety
-- **NVCF**: microsoft/trellis text-to-3D (the free default forge lane) and FLUX.1-schnell text-to-image
+- **NVCF**: microsoft/trellis text-to-3D (the free default forge lane) and FLUX.1-dev text-to-image (the hosted FLUX.1-schnell preview was retired, so do not claim schnell)
- **NVIDIA Riva**: Magpie multilingual TTS and Riva ASR over NVCF gRPC
- **NVIDIA Audio2Face-3D**: audio-driven ARKit blendshape facial animation, live at three.ws/demos/audio2face
- **RTX PRO 6000 Blackwell**: validated (sm_120 builds) for the heavy image-to-3D lane; not the default production GPU
diff --git a/docs/nvidia-apps-catalog-request.md b/docs/nvidia-apps-catalog-request.md
index fa43b6809..b76cde18c 100644
--- a/docs/nvidia-apps-catalog-request.md
+++ b/docs/nvidia-apps-catalog-request.md
@@ -1,15 +1,39 @@
-# The NVIDIA Inception email: five asks, one thread
+# The NVIDIA Inception email: send-ready packet
-The Inception portal has no "publish to the catalog" control. A product record makes the company eligible; the public [Accelerated Apps Catalog](https://marketplace.nvidia.com/en-us/enterprise/applications/) is curated and published by NVIDIA. This is the request that asks for the listing, sent after the portal record is corrected per [nvidia-apps-catalog-listing.md](./nvidia-apps-catalog-listing.md).
+Everything needed to send the Accelerated Apps Catalog request in one sitting: the pre-send portal fix, the paste-ready email, the assets, and what to do with each reply.
-It also carries four other asks that have never been made. Per [nvidia-visibility-map.md](./nvidia-visibility-map.md), every Tier 1 ask routes to the same inbox, so they are batched here deliberately. **Do not split these into separate emails.** A program manager who receives one clear email with five asks answers it. Five separate emails read as noise.
+The Inception portal has no "publish to the catalog" control. A product record makes the company eligible; the public [Accelerated Apps Catalog](https://marketplace.nvidia.com/en-us/enterprise/applications/) is curated and published by NVIDIA. This email is the request that asks for the listing, and it carries four other asks that have never been made. Per [nvidia-visibility-map.md](./nvidia-visibility-map.md), every Tier 1 ask routes to the same inbox, so they are batched here deliberately. **Do not split these into separate emails.** A program manager who receives one clear email with five asks answers it. Five separate emails read as noise.
- **Subject:** three.ws (Inception member): catalog listing, Showcase, and co-marketing
+- **From:** the address on the Inception portal account
- **Sent:** not yet sent. Record the date here when it goes out.
-- **Blocked on:** the portal record correction, [nvidia-apps-catalog-listing.md](./nvidia-apps-catalog-listing.md) step 1. As filed the record reads as a CUDA consumer rather than a digital-human app, which undercuts asks 1 and 2. Correct it, save it, then send.
-### The five asks, and why each one is here
+---
+
+## Step 1: fix the portal record first (about 5 minutes)
+
+Do this before sending. The email's first two asks are judged against the record, and as filed the record reads as a CUDA consumer rather than a digital-human app. Verified 2026-09-04: still uncorrected.
+
+Go to the Inception portal, Products, then Edit on the `three.ws` row.
+
+**Move these from Considering to Technologies Used:**
+
+| Toggle | Evidence it is shipping |
+|---|---|
+| **Riva** | [api/_lib/asr-nvidia.js](../api/_lib/asr-nvidia.js) (ASR) and [api/_lib/tts-nvidia.js](../api/_lib/tts-nvidia.js) (Magpie TTS), both Riva gRPC over NVCF |
+| **Audio2Face** | [api/_lib/a2f-nvidia.js](../api/_lib/a2f-nvidia.js) against `grpc.nvcf.nvidia.com`, browser playback in [src/voice/a2f-player.js](../src/voice/a2f-player.js), live at [/demos/audio2face](https://three.ws/demos/audio2face) |
+| **NIM microservices** (if the form offers it as hosted APIs) | `integrate.api.nvidia.com`: Nemotron 3 Super and Nano, Nemotron Nano VL, NV-EmbedQA-E5-v5, NemoGuard 8B |
+
+**Uncheck entirely: DeepVariant NIM.** It is genomics variant calling. It has nothing to do with this product and it weakens category fit with a curator who reads the record.
+
+**Leave under Considering: TensorRT, Triton Inference Server, Omniverse Kit.** The GPU workers are PyTorch plus CUDA today and OpenUSD interop is roadmap. Do not claim them.
+
+**Keep under Used, and add if the form offers them:** CUDA Toolkit, cuDNN, cuBLAS, CUDA Python, NVIDIA Kaolin, nvdiffrast, L4 GPUs, RTX PRO 6000 Blackwell.
+
+Save the record. Then send the email below.
+
+## Step 2: the five asks, and why each one is in this email
| # | Ask | Basis |
|---|---|---|
@@ -21,65 +45,171 @@ It also carries four other asks that have never been made. Per [nvidia-visibilit
Asks 2 and 3 are entitlements listed on NVIDIA's own [program page](https://www.nvidia.com/en-us/startups/), not favors. Ask 3 is the specific unlock for social amplification: NVIDIA's social team can only reshare a post carrying the official badge and `#NVIDIAInception`. Without the kit there is no mechanism for them to amplify anything.
----
+## Step 3: paste this
+
+Plain text, no formatting needed. Every link in it was checked and answered 200 on 2026-09-04.
+```
Hello,
-three.ws is an NVIDIA Inception member (accepted July 2026). Our product record is filed in the portal as Shipping / GPU Accelerated. I have five requests, batched into one email rather than five.
+three.ws is an NVIDIA Inception member (accepted July 2026). Our product
+record is filed in the portal as Shipping / GPU Accelerated. I have five
+requests, batched into one email rather than five.
-## 1. Accelerated Apps Catalog
+
+1. ACCELERATED APPS CATALOG
I would like to request consideration for the Accelerated Apps Catalog.
-**What it is:** three.ws is a browser-native platform for generating 3D avatars and AI agents from a prompt. Text, image, or sketch input becomes a textured, auto-rigged GLB in seconds, and the resulting agent can talk, lip-sync, and be embedded on any site with one script tag. Free text-to-3D at https://three.ws/forge, no signup.
+What it is: three.ws is a browser-native platform for generating 3D avatars
+and AI agents from a prompt. Text, image, or sketch input becomes a textured,
+auto-rigged GLB in seconds, and the resulting agent can talk, lip-sync, and be
+embedded on any site with one script tag. Free text-to-3D at
+https://three.ws/forge, no signup.
+
+Why it belongs in the catalog: the whole product is NVIDIA accelerated
+computing end to end, in two layers.
+
+Self-hosted: eight GPU workers on Google Cloud Run, twelve service deployments
+across two regions, eleven on NVIDIA L4 and one on RTX PRO 6000 Blackwell.
+They are built on nvidia/cuda 12.1 to 12.8 images with custom-compiled CUDA
+extensions (nvdiffrast, diffoctreerast, diff-gaussian-rasterization,
+torchmcubes) and NVIDIA Kaolin, and they run Hunyuan3D 2.1 and TRELLIS for
+image-to-3D, TripoSG for sketch-to-3D, TripoSR for fast drafts,
+Make-It-Animatable for auto-rigging, and Motion Diffusion Model for
+text-to-motion. The heavy image-to-3D lane is validated on RTX PRO 6000
+Blackwell with sm_120 kernel builds.
+
+NVIDIA-hosted: TRELLIS on NVCF is our default free text-to-3D lane. Nemotron 3
+Super and Nano power agent reasoning, Nemotron Nano VL handles vision,
+NV-EmbedQA-E5-v5 and Mistral reranking power agent memory, NemoGuard 8B gates
+every public publish, Riva Magpie TTS and Riva ASR give agents a voice, and
+Audio2Face-3D drives real-time ARKit blendshape facial animation in the
+browser (live demo: https://three.ws/demos/audio2face).
-**Why it belongs in the catalog:** the whole product is NVIDIA accelerated computing end to end, in two layers.
+Public engineering work on the stack:
-Self-hosted: eight GPU workers on Google Cloud Run, twelve service deployments across two regions, eleven on NVIDIA L4 and one on RTX PRO 6000 Blackwell. They are built on `nvidia/cuda` 12.1 to 12.8 images with custom-compiled CUDA extensions (nvdiffrast, diffoctreerast, diff-gaussian-rasterization, torchmcubes) and NVIDIA Kaolin, and they run Hunyuan3D 2.1 and TRELLIS for image-to-3D, TripoSG for sketch-to-3D, TripoSR for fast drafts, Make-It-Animatable for auto-rigging, and Motion Diffusion Model for text-to-motion. The heavy image-to-3D lane is validated on RTX PRO 6000 Blackwell with `sm_120` kernel builds.
+- Image-to-3D on NVIDIA L4 and Blackwell (memory ceilings, sm_120 kernels,
+ regional GPU quota):
+ https://three.ws/blog/image-to-3d-on-nvidia-l4-and-blackwell
+- How Nemotron made our text-to-3D pipeline usable, posted to the NVIDIA
+ Developer Forums:
+ https://forums.developer.nvidia.com/t/how-nemotron-made-three-ws-text-to-3d-pipeline-usable/376445
+- How we translate the app into 100 languages with NVIDIA NIM, also on the
+ Developer Forums:
+ https://forums.developer.nvidia.com/t/how-three-ws-translates-a-web-app-into-100-languages-with-nvidia-nim-an-llm-powered-i18n-pipeline/377379
+- Our full NVIDIA integration map, model by model:
+ https://three.ws/docs/nvidia-models
-NVIDIA-hosted: Nemotron Super and Nano power agent reasoning, Nemotron Nano VL handles vision, NV-EmbedQA-E5-v5 and Mistral reranking power agent memory, NemoGuard 8B gates every public publish, Riva Magpie TTS and Riva ASR give agents a voice, and Audio2Face-3D drives real-time ARKit blendshape facial animation in the browser (live demo: https://three.ws/demos/audio2face).
+Suggested categorization: Industry, Media and Entertainment. Workload,
+Generative AI and Digital Humans. Type, SaaS web application.
-**Public engineering work on the stack:**
+If anything else is needed for the listing (logo, screenshots, a longer
+description, a specific form), tell me the format and I will send it the same
+day. Brand assets and a press kit are ready at
+https://three.ws/brand/three-ws-press-kit.zip and I can supply 1920x1080
+captures of the generation flow on request.
-- Image-to-3D on NVIDIA L4 and Blackwell (memory ceilings, `sm_120` kernels, regional GPU quota): https://three.ws/blog/image-to-3d-on-nvidia-l4-and-blackwell
-- How Nemotron made our text-to-3D pipeline usable, posted to the NVIDIA Developer Forums: https://forums.developer.nvidia.com/t/how-nemotron-made-three-ws-text-to-3d-pipeline-usable/376445
-- Our full NVIDIA integration map, model by model: https://three.ws/docs/nvidia-models
-Suggested categorization: Industry, Media and Entertainment. Workload, Generative AI and Digital Humans. Type, SaaS web application.
+2. INCEPTION STARTUP SHOWCASE
-If anything else is needed for the listing (logo, screenshots, a longer description, a specific form), tell me the format and I will send it the same day. Press kit and brand assets are ready at https://three.ws and I can supply 1920x1080 captures of the generation flow on request.
+The program page describes member spotlights and story opportunities through
+the Inception Startup Showcase. I would like to put three.ws forward for
+consideration.
-## 2. Inception Startup Showcase
+The story we would tell is the one in the engineering posts above: a small team
+running a self-hosted NVIDIA GPU fleet plus NVIDIA-hosted models to make 3D
+agent generation fast enough and cheap enough to give away for free in a
+browser, with no signup. If a written profile, a founder interview, or a
+specific template is the format you need, tell me which and I will turn it
+around the same week.
-The program page describes member spotlights and story opportunities through the Inception Startup Showcase. I would like to put three.ws forward for consideration.
-The story we would tell is the one in the engineering posts above: a small team running a self-hosted NVIDIA GPU fleet plus NVIDIA-hosted models to make 3D agent generation fast enough and cheap enough to give away for free in a browser, with no signup. If a written profile, a founder interview, or a specific template is the format you need, tell me which and I will turn it around the same week.
+3. CO-MARKETING ASSETS
-## 3. Co-marketing assets
+The program page lists official badges, co-branded social content, and
+customizable event assets as a member benefit. We have never requested the kit,
+and as a result our Inception membership has never been announced on our own
+channels, because we would rather announce it correctly the first time.
-The program page lists official badges, co-branded social content, and customizable event assets as a member benefit. We have never requested the kit, and as a result our Inception membership has never been announced on our own channels, because we would rather announce it correctly the first time.
+Please send the current co-marketing kit and the badge usage guidelines. I want
+to confirm the correct badge lockup, the required attribution wording, and
+which handle to tag for digital-human content before we post anything.
-Please send the current co-marketing kit and the badge usage guidelines. I want to confirm the correct badge lockup, the required attribution wording, and which handle to tag for digital-human content before we post anything.
-## 4. A redirect to the ACE / digital-human team
+4. A REDIRECT TO THE ACE / DIGITAL-HUMAN TEAM
-I could not find a public intake route for the ACE team, so I am asking here for a redirect.
+I could not find a public intake route for the ACE team, so I am asking here
+for a redirect.
-We are running Audio2Face-3D in a browser tab. The client streams 16 kHz mono PCM into the ACE `A2FControllerService` over NVCF gRPC and renders the returned ARKit blendshape track against a WebGL avatar, mapping ARKit-52 onto whatever morph convention the user's own model ships, alongside Riva ASR in and Magpie TTS out. Live and public, no install and no signup: https://three.ws/demos/audio2face
+We are running Audio2Face-3D in a browser tab. The client streams 16 kHz mono
+16-bit PCM into the ACE A2FControllerService over NVCF gRPC and renders the
+returned ARKit blendshape track against a WebGL avatar, mapping ARKit-52 onto
+whatever morph convention the user's own model ships (ARKit, VRM vowels,
+Oculus visemes), alongside Riva ASR in and Magpie TTS out. Live and public, no
+install and no signup: https://three.ws/demos/audio2face
-To be precise about what we are claiming: we consume ACE microservices as hosted APIs. We do not self-host NIM. Most A2F work we have seen lives in Unreal or Omniverse, so a browser-native implementation seemed worth putting in front of the team that owns it, whether that is useful as a reference integration, a feedback channel, or neither.
+To be precise about what we are claiming: we consume ACE microservices as
+hosted APIs. We do not self-host NIM. Most A2F work we have seen lives in
+Unreal or Omniverse, so a browser-native implementation seemed worth putting in
+front of the team that owns it, whether that is useful as a reference
+integration, a feedback channel, or neither.
-## 5. GTC Startup Pavilion and other events
-I understand the GTC 2026 call for submissions has closed and that Pavilion placement is arranged with the program team rather than through the CFP. Two questions:
+5. GTC STARTUP PAVILION AND OTHER EVENTS
-- Is there a route to the Inception Startup Pavilion for the next GTC, and when does that conversation normally start?
-- The program materials mention Inception exposure at Oracle AI World, KubeCon, Supercomputing, Microsoft Ignite, AWS re:Invent, and NeurIPS. Which of those carry startup slots we would be eligible for, and what are their lead times?
+I understand the GTC 2026 call for submissions has closed, and that Pavilion
+placement is arranged with the program team rather than through the CFP. Two
+questions:
+
+- Is there a route to the Inception Startup Pavilion for the next GTC in San
+ Jose, and when does that conversation normally start? I would also like to
+ know when the call for submissions is expected to open so we do not miss it.
+- The program materials mention Inception exposure at Oracle AI World, KubeCon,
+ Supercomputing, Microsoft Ignite, AWS re:Invent, and NeurIPS. Which of those
+ carry startup slots we would be eligible for, and what are their lead times?
+
+I would rather ask now and plan around the real dates than miss a window by a
+month.
-I would rather ask now and plan around the real dates than miss a window by a month.
Thank you,
Nicholas
three.ws
https://three.ws
+```
+
+## Attachments
+
+None are required; every asset is linked in the body and the links are live. Attach only if you prefer the curator not to have to click:
+
+- Logo mark: [public/brand/three-ws-mark.png](../public/brand/three-ws-mark.png)
+- Lockups: [three-ws-lockup-on-dark.png](../public/brand/three-ws-lockup-on-dark.png), [three-ws-lockup-on-light.png](../public/brand/three-ws-lockup-on-light.png)
+- Press kit (6.3 MB, may trip attachment limits; the link in the body is safer): [public/brand/three-ws-press-kit.zip](../public/brand/three-ws-press-kit.zip)
+
+Screenshots are deliberately not attached. The email offers them on request, which is a cheaper second touch than guessing the spec now. If they ask, capture 1920x1080 from [/forge](https://three.ws/forge) (generation in progress plus result), [/agora](https://three.ws/agora), and [/character-library](https://three.ws/character-library).
+
+## After you send
+
+1. Record the send date in the **Sent** field at the top of this doc.
+2. File the second product record, the `<agent-3d>` digital-human embed. Copy is ready in [nvidia-apps-catalog-listing.md](./nvidia-apps-catalog-listing.md) under "Second product record". It targets the Digital Humans workload filter that the main record does not, and it is independent of the reply.
+3. Route each reply:
+
+| Reply | Where it goes next |
+|---|---|
+| Catalog form or asset request | Answer same day. Everything is in [nvidia-apps-catalog-listing.md](./nvidia-apps-catalog-listing.md). |
+| Showcase interest | Draft the profile against the engineering posts, not marketing copy. |
+| Co-marketing kit arrives | Unblocks the badge post. Copy is drafted in [marketing/nvidia-inception/social-copy.md](../marketing/nvidia-inception/social-copy.md), owner-gated. |
+| ACE redirect | The forum post 3 draft, [nvidia-forum-browser-digital-human.md](./nvidia-forum-browser-digital-human.md), is the technical brief to send them. |
+| GTC dates | Put them on the watch in [nvidia-visibility-map.md](./nvidia-visibility-map.md) Tier 3 and write the abstracts. |
+| No reply in 3 weeks | One follow-up on ask 1 only, and use the "Maximize Your Membership" webinar Q&A to ask 2 and 3 live. |
+
+## Verification log
+
+Checked 2026-09-04 before this packet was finalized:
+
+- Every URL in the email body returned HTTP 200, including both Developer Forums posts, the blog post, the demo, the docs pages, and the press kit.
+- The Riva, Audio2Face, and NIM claims were re-read against the code they cite. A2F still targets `nvidia_ace.services.a2f_controller.v1.A2FControllerService` on `grpc.nvcf.nvidia.com:443` at 16 kHz mono, matching the wording in ask 4.
+- The text-to-image lane is FLUX.1-dev, not schnell (the hosted schnell preview was retired). FLUX is not claimed in this email, so no change was needed here.
+- The GPU fleet count (eight workers, twelve deployments, eleven L4, one RTX PRO 6000) is carried over from the 2026-09-02 live recount. `gcloud` credentials in the dev container had expired, so it was not re-run today. To refresh before sending: `gcloud run services list --project aerial-vehicle-466722-p5 --format="csv[no-heading](metadata.name,spec.template.spec.nodeSelector)" | grep -i accelerator`