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.
NVIDIA visibility map: every surface, and how to get on it
Inception membership is the door, not the result. This doc is the complete map of every NVIDIA surface where three.ws can earn recognition, what each one actually requires, and the honest status of each today.
Scope note: the Accelerated Apps Catalog already has its own two docs and is not re-litigated here. See nvidia-apps-catalog-listing.md for the listing kit and nvidia-apps-catalog-request.md for the inclusion email. This doc covers everything else and links to them where they overlap.
Researched and verified 2026-08-17. Every intake URL and email below was fetched or searched the same day; the ones that turned out to be dead are recorded in Dead ends so nobody re-chases them.
Program-team update, 2026-09-14. NVIDIA answered the consolidated request. Catalog consideration runs through periodically reviewed Shipping product records with complete runtime-technology evidence, descriptions, technical details, logo, and brand color. The Co-Branded Marketing Assets benefit is accessible again. The GTC Startup Pavilion is invite-only and selected from current member profiles, demos, and relationships. Startup Showcase routing and an ACE introduction were not directly answered, so those are the only two questions worth sending back.
The two facts that should reorder priorities
Before any new outreach, two things about our existing position:
1. The Inception membership was never announced on X. announcement-coverage.md
tracks an "announced on X" column per surface. /nvidia (added 2026-07-30) has
that column empty, while /openai has a real post behind it. The single biggest
NVIDIA credential we hold has had no social push on our own channels. Everything
in the Social section is blocked on nothing and
costs nothing.
2. Jensen Huang already interacted with three.ws content, and we have no
receipt. data/timeline.json carries
2026-07-18-jensen-huang: "NVIDIA CEO Jensen Huang interacts with three.ws
content, amplifying the NVIDIA Inception announcement." Its source_url is
null. Four of the seven NVIDIA timeline events have a null source_url,
including the Inception acceptance and its public announcement. (The Developer
Program event now cites the 2026-07-11 @trythreews acceptance post.)
The repo copy of the X archive cannot close this gap: on 2026-09-16 every
NVIDIA-related post in data/x-archive/trythreews-2026-08-14.json was checked
and none records the interaction, and no Inception announcement post exists
there either. An engagement from another account shows up only in the
account's own notifications or the other account's activity, so the receipt
has to come from whoever holds the @trythreews or @nichxbt login.
That is the highest-value social proof this company has and it is currently
unciteable. An engagement from NVIDIA's CEO is the first line of any Showcase
pitch, podcast nomination, or press approach, and right now it cannot be shown to
anyone. Recovering that link is the highest-leverage single action on this
page, and it is a five-minute job for whoever has the account history. Archive
it to web.archive.org the moment it is found, because a deleted post takes the
proof with it.
Tier 1: act now, no dependencies
Everything here can be done today. Nothing waits on NVIDIA, a build, or a date.
| # | Surface | Intake | Why it is Tier 1 |
|---|---|---|---|
| 1 | Our own X, LinkedIn, and Telegram | Ours | The membership post that never happened. See Social. |
| 2 | Inception co-marketing kit | Portal > Benefits > Co-Branded Marketing Assets |
NVIDIA confirmed on 2026-09-14 that the page is back up. Download the current package, reconcile its badge/guidelines, then publish. |
| 3 | Portal product record correction | programs.nvidia.com/phoenix/application |
Gates the Showcase AND the Apps Catalog. Already specified in nvidia-apps-catalog-listing.md step 1. Nothing else in this doc lands well while the record says "CUDA consumer". |
| 4 | Inception Startup Showcase | Existing [email protected] thread |
Requested 2026-09-04 and not directly answered; re-asked in the reply sent 2026-10-25. After #3, send only the focused routing question drafted in nvidia-apps-catalog-request.md. |
| 5 | AI Podcast guest nomination | Embedded form at ai-podcast.nvidia.com | Free, one form, and a founder-shaped story. Answers drafted below: AI Podcast nomination. Re-verified 2026-09-04: the form is live and self-nomination is explicitly invited. |
| 6 | NVIDIA Developer Forums, post 5 | forums.developer.nvidia.com | We are 4 for 4 on posts that landed. This is the one channel with a proven track record, and it is self-serve. The Audio2Face post went live 2026-09-22. Next up: nvidia-forum-gpu-fleet-post.md, cut to the approval pattern. Owner-gated on the external-channel rule, then paste and post. See Technical publishing. |
| 7 | NVIDIA Developer Discord | discord.com/invite/nvidiadeveloper | NVIDIA engineers and product staff are in there. The browser Audio2Face demo is a conversation starter with the actual ACE team, not a marketing ask. |
| 8 | Fix docs/listings.md |
Ours | The "Startup & Credit Programs" table lists Quicknode and Google Cloud and has no NVIDIA row at all, which is the one canonical place a reader looks for our program memberships. Fixed in this change. |
| 9 | "Maximize Your Membership" benefits webinar | Region registration from the Inception member mailer (Americas / EMEA / APAC) | Announced to members 2026-09-02. NVIDIA's own walkthrough of requesting and managing benefits through the portal, which is the exact mechanism gating items 2, 3 and 4 on this list. The batched email went out 2026-09-04, so use the Q&A to chase asks 2 and 3 live rather than waiting on the inbox. |
AI Podcast nomination, paste-ready
The form is embedded at ai-podcast.nvidia.com and its fields were re-read 2026-09-04. Self-nomination is invited in NVIDIA's own copy ("Have a podcast guest proposal for you or someone you know?"). One submission, no gatekeeper, and it is independent of the catalog email.
| Field | Value |
|---|---|
| First name | Nicholas |
| Last name | Yours |
| Business email | [email protected] (the address the Inception thread already runs on) |
| Organization | three.ws |
| Industry | Media and Entertainment if offered, otherwise Software and Internet |
| Job title | Founder |
| Location | Your country of residence. The form reveals a State/Province field conditionally, so answer that one only if it appears. |
| Preferred language | English |
Description field:
three.ws is a browser-native platform that turns a sentence into a rigged,
animated 3D AI agent. Type a prompt and a textured, auto-rigged 3D character
appears in seconds; it then talks, lip-syncs, and embeds on any website with a
single script tag. The free lane requires no signup and no install.
The episode I would want to record is about what it takes to give that away for
free. We run a self-hosted fleet of NVIDIA L4 GPU workers alongside
NVIDIA-hosted models: TRELLIS and Hunyuan3D for generation, Nemotron for agent
reasoning, Riva for voice, and Audio2Face-3D for facial animation streaming
into a browser tab rather than into Unreal or Omniverse. The engineering is
public, including our field notes on running image-to-3D on L4 and Blackwell
with sm_120 kernel builds, and two posts on the NVIDIA Developer Forums.
three.ws is an NVIDIA Inception member. Happy to bring a live demo: the fastest
version of this story is watching a face lip-sync in a browser tab with nothing
installed.
Do not attach anything. The form has no upload field; the demo links belong in the description only if the field accepts URLs, and it does, so the two worth including if you want them are https://three.ws/forge and https://three.ws/demos/audio2face.
Tier 2: needs a build first
Real listings, real effort, honest gates. Do not pitch these before the code exists.
| Surface | Intake | The gate |
|---|---|---|
| NGC Catalog (via the NGC Software Partner program) | A self-serve "Become an NGC Software Partner" form on that page, then the partner legal agreement, a push to an NGC private staging repo, security scanning and QA, final sign-off | The closest Tier 2 surface to reachable, and the only NVIDIA directory with a self-serve door. It now has its own kit: nvidia-ngc-listing.md audits all four prerequisites against the code, names the publishable container (model-trellis, MIT upstream, no baked weights, no telemetry), and carries the paste-ready listing copy. The EULA prerequisite is closed (/legal/nvidia-ngc-eula). What remains is one Cloud Build run to make the image runnable outside our GCP project, plus the owner-signed legal agreement. |
| ACE / digital-human ecosystem | No public form found. Ask for the redirect in the Showcase email | Genuinely strong: Riva, Magpie, and Audio2Face-3D are shipping, and a browser-native A2F demo is rare (most A2F work lives in Unreal or Omniverse). This is our most differentiated technical story. Do not overclaim: we consume ACE microservices, we do not self-host NIM. |
| NVIDIA Technical Blog (developer.nvidia.com/blog) | No public guest-submission process. Route via the forum Technical Blog category and the program team | Editorially controlled by NVIDIA. The realistic path is a forum post that performs, then a pitch. Our two live forum posts are the audition tape. |
| Omniverse Exchange | Omniverse Exchange Publishing Portal (early access), plus a GitHub repo tagged with the omniverse-kit-extension topic |
Blocked and should stay blocked. OpenUSD interop is roadmap, not shipping, and nvidia-apps-catalog-listing.md already rules that Omniverse Kit stays under "Considering". Note the Omniverse Launcher was deprecated 2025-10-01, so old guides mislead. |
Tier 3: time-gated, and the clock is running
GTC is the single largest recognition surface NVIDIA operates, and the window for the next one is open approximately now.
GTC 2026 (San Jose, March 16 to 19) ran an Inception Startup Pavilion with 55+ startups, "NVIDIA Inception Presents: Startup Pitches" sessions, and 25 Startup Spotlight sessions. NVIDIA's own startup page for the event is nvidia.com/gtc/startups.
Status as of 2026-08-17: the call for submissions page reads "The Poster Call for Submissions and Content Interest Survey are now closed" for GTC 2026. NVIDIA's own FAQ page is already titled for GTC San Jose 2027, and third-party event trackers report 2027-03-14 to 2027-03-18. Treat the dates as unconfirmed until NVIDIA states them, but treat the submission window as imminent: for a March event, recent cycles opened submissions in the summer and closed in early fall, which is the month we are in.
Re-checked 2026-09-02. Still closed: the call-for-submissions page carries the same GTC 2026 "now closed" copy, and no 2027 CFP or registration has opened. Third-party trackers have converged on 2027-03-15 to 2027-03-18, San Jose, one day later than the dates recorded above, which is a reminder that none of these dates are NVIDIA-stated yet. Nothing to submit to this week. Keep the weekly watch running; the abstracts in step 2 below are what should be written while the window is shut.
Re-checked 2026-10-08: the window is open. NVIDIA now states the dates, 2027-03-15 to 03-18, San Jose. The poster call closes 2026-11-10 at 5 p.m. PT and the Content Interest Survey closes 2026-10-09 at 5 p.m. PT. No open speaker call is posted. Both poster submissions and the survey answers are written in nvidia-gtc-2027-poster.md; what remains is the owner submitting them as the presenting author.
Actions, in order:
- Put nvidia.com/gtc/call-for-submissions on a weekly watch. Missing this window costs a full year. It is the one item on this page with a hard, externally-set deadline.
- Write the poster and session abstracts before the form opens. Done 2026-10-08: nvidia-gtc-2027-poster.md. The strongest submission we have is the browser-native digital human: Audio2Face-3D plus Riva plus Nemotron driving a rigged avatar in a tab with no install, on a fleet of L4s and one Blackwell. Second-strongest is the L4-and-Blackwell engineering material already published on our blog.
- Do not search for a Startup Pavilion form. NVIDIA confirmed on 2026-09-14 that the Pavilion is exclusive and invite-only. Keep the Shipping product record complete, keep the browser Audio2Face demo reliable, and send real technical proof through the existing relationship; those are the selection signals NVIDIA named.
- Do not forget the Content Interest Survey. It closes alongside the CFP and is a low-effort way to signal topic fit before sessions are locked.
Also date-driven, lower stakes: NVIDIA lists Inception exposure at Oracle AI World, KubeCon, Supercomputing, Microsoft Ignite, AWS re:Invent, and NeurIPS. Worth asking which of those have startup slots we qualify for.
Social surfaces (ours to drive)
This is the section with no gatekeeper, and it is the one we have used least.
The mechanics that actually get amplified: NVIDIA's Inception co-marketing
benefit is explicitly "official badges, co-branded social content, and
customizable assets for events". Posts that use the official badge, tag the right
account, and carry #NVIDIAInception are the ones NVIDIA's social team can
reshare. A generic post naming NVIDIA in prose is not resharable.
Pick the account by topic rather than blasting the main handle. Verified from NVIDIA's official social directory:
| Our content | Account to tag | Platform |
|---|---|---|
| Digital humans, Audio2Face, Riva, avatars | @NVIDIAAIDev, @NVIDIADeveloper |
X |
| Inception membership, founder story, milestones | NVIDIA for Startups (82k+ followers) | |
| Nemotron, NIM, inference pipeline | @NVIDIAAI, @NVIDIAAIDev |
X |
| 3D, OpenUSD, scene work | @NVIDIAOmniverse |
X, LinkedIn, Discord |
| GTC submissions and attendance | @NVIDIAGTC |
X |
| General company news | @NVIDIA, @NVIDIANewsroom |
X |
Also live and worth using: NVIDIA Developer on YouTube, Instagram, and Threads, plus the two Discords (NVIDIA Developer, NVIDIA Omniverse).
The content we already have and have never posted:
- The membership itself, with the official badge. Never posted.
- The browser Audio2Face demo at
/demos/audio2face. A 15-second screen capture of a face lip-syncing in a browser tab with no install is the most screenshot-friendly asset on the platform, and it maps exactly to what the ACE team promotes. - The two NVIDIA Developer Forums posts. Published on NVIDIA's own property and never amplified from ours.
- The L4-and-Blackwell engineering post, including the
sm_120kernel and regional-quota material that GPU engineers actually reply to. - readme-3d: an interactive 3D model rendered natively in a GitHub README. Genuinely novel, and the model in it was generated on the NVIDIA lane.
Sequence it: badge post, then demo clip, then the forum posts, then the engineering deep dive. One per week beats four in a day, and each one is a separate chance for a reshare.
Technical publishing (the proven channel)
Four forum posts, all live, all landed:
- How Nemotron made three.ws text-to-3D pipeline usable
- How three.ws translates a web app into 100 languages with NVIDIA NIM
- NVIDIA NIM model retirements: what a 410 Gone does to a fallback chain (posted 2026-09-21 from nvidia-forum-model-retirement-post.md, written to the pattern below)
- A digital human in a browser tab: streaming Audio2Face-3D onto whatever rig the visitor brought (posted 2026-09-22 from nvidia-forum-browser-digital-human.md, revised to the pattern the same day)
This is the only NVIDIA channel where we have a track record, it is self-serve, and it feeds everything else: forum posts are what a Technical Blog pitch and a GTC abstract are built on.
What the approved posts have in common (read before drafting the next one)
Three topics from the nichxbt account had cleared moderation as of 2026-09-21: the
first two above and a CloudXR 6.2 support question, which drew an NVIDIA employee
reply the same day. We cannot see the forum's moderation settings, so this is
the pattern in what passed, plus the stock Discourse new-user defaults (two
links and one image per post, and an automatic hold once a new account keeps
linking to the same external host).
- Filed under a product category with the forum's own tags. Both articles
sit in AI & Data Science > NVIDIA NIM > Models, tagged from the existing
vocabulary (
nim,llama,nemotron). The "suggested tags" we wrote at the bottom of the i18n post (localization,i18n) do not exist there and were never applied. There is no "community showcase" category to aim at. - An NVIDIA product in the title, and exact model ids in the body.
- Almost no hyperlinks. The Nemotron post carries one link and the i18n
post none.
three.wsand repo paths such asdocs/nvidia-models.mdappear as plain text. The two unposted drafts carry five and ten links, mostly to our own domain and our own GitHub, which is the shape a spam hold looks for. Cut them to two or fewer before posting, and prefer links back into the forum itself. - No launch voice and no commerce vocabulary. Each opens on a technical situation in the first person. Nothing about tokens, pricing, or monetization beyond a config line.
- Real artifacts. A prompt quoted from the source file, a measured token count, a median latency with its tail. At most one image.
- Something for NVIDIA to answer. The Nemotron post closes with three concrete product asks, the CloudXR post with a direct question. A post a staff member can usefully reply to is a post a moderator wants live.
- Length of a four to five minute read, roughly 1,500 to 2,300 words.
Post 3 should be the browser digital human, because it is the piece nobody else has written: streaming Audio2Face-3D blendshape tracks over gRPC into a WebGL avatar, mapping ARKit-52 onto whatever morph convention the visitor's model ships (ARKit, VRM vowels, Oculus visemes), with Riva ASR in and Magpie TTS out. Write it against api/_lib/a2f-nvidia.js and src/voice/a2f-player.js so every claim has code behind it, and follow the house rule the existing two follow: real numbers, real latencies, no marketing voice.
It is live: topic 383953, from nvidia-forum-browser-digital-human.md (drafted 2026-09-02, posted 2026-09-22). It covers the NVCF bidirectional stream, the 44.1 kHz to 16 kHz resampling contract and why playback must use the original audio, the derived blendshape path that lets a VRM or Oculus-viseme rig lip-sync from an ARKit-52 track, the double-stacking and 30-to-60 fps interpolation bugs that only appear on real uploads, and the function-id rotation that took the lane down once. It was revised on 2026-09-22 to the approval pattern above: two links (the live demo and our retirement topic), plain-text repo paths, the NIM > Models category, latencies re-measured against production that day, and the NemoGuard claim removed (it screens outbound publishing, not chat). It is owner-gated under the external-channel rule and cleared moderation the day it was posted.
Drafted, not yet posted
| Draft | Subject |
|---|---|
| nvidia-forum-gpu-fleet-post.md | The production GPU fleet: cold weight loads on L4s, min-instances as a quota decision rather than a performance one, the keep-warm cron, and why a failover chain must never have an empty rung. Cut to the approval pattern 2026-09-22 (two links, both to our own topics here); ready to paste |
Member benefits and current disposition
Straight from NVIDIA's Inception program page. These are membership benefits, not favors, and they are requested through the portal or the program email.
| Benefit | NVIDIA's wording | Status |
|---|---|---|
| Co-marketing assets | "Amplify your brand with official badges, co-branded social content, and customizable assets for events." | Access restored 2026-09-14. Owner downloads from Portal > Benefits > Co-Branded Marketing Assets and reconciles the current kit before posting. |
| Startup Showcase feature | "Feature in the NVIDIA Inception Startup Showcase with member spotlights and story opportunities." | Requested 2026-09-04; NVIDIA's reply did not address routing. One focused question remains. |
| Member-only newsletter | Product releases and offers | Confirm we are subscribed. Also the likeliest place a member spotlight surfaces. |
| Event exposure | GTC, Oracle AI World, KubeCon, Supercomputing, Microsoft Ignite, AWS re:Invent, NeurIPS | Never asked which have startup slots. |
| Capital Connect / VC Alliance | Eligibility-based investor introductions | Program page. Owner's call, not an engineering task. |
| Cloud credits and preferred pricing | Free credits from NVIDIA and partners | Separate from the GCP credits already in use. Worth auditing against ops/gcp-credits-plan.md. |
On tiers: third-party write-ups describe a Premier tier above entry with a larger DGX allocation and more co-marketing, promoted on usage and traction signals. NVIDIA does not document this publicly, so treat it as unconfirmed and never claim a tier. It is a fair question to ask the program team, and the answer is worth having, because "more co-marketing" is exactly what this page is about.
Dead ends: do not spend time here
Recorded so the next person does not repeat the search.
- NVIDIA Connect for ISVs.
nvidia.com/en-us/programs/isv/301-redirects todeveloper.nvidia.com. The standalone ISV program has been folded into the Developer Program. There is no separate ISV listing to apply for. - A self-serve "list my product" form, on the marketing surfaces. None
exists for the Apps Catalog, the Showcase, or any other surface on this page:
each runs through the portal record plus a human at
[email protected]. The single exception is NGC, which does publish a partner intake form; see nvidia-ngc-listing.md. Stop looking for a form anywhere else. - Omniverse Launcher guides. Deprecated 2025-10-01. Any tutorial routing through it is stale.
- build.nvidia.com as a publishing target. It is where we get our
nvapi-key and consume hosted models. Publishing goes through NGC, not there.
Contacts, in one place
| Purpose | Address |
|---|---|
| All Inception matters, Showcase, co-marketing, GTC pavilion | [email protected] |
| China-based startups (not us, recorded for completeness) | [email protected] |
| Member portal | programs.nvidia.com/phoenix/application |
| Live phone support (Mon to Fri, 8:00 to 17:00 PT) | +1 (408) 486-2056 |
| Program FAQ and benefits webinar registration | Linked from the Inception member mailer; region-specific (Americas, EMEA, APAC) |
One thread, one address. Every ask in Tier 1 goes to the same inbox, so batch them into a single email rather than sending four.
Sent 2026-09-04; answered 2026-09-14. nvidia-apps-catalog-request.md carries all five asks and the reply disposition. Do not start a second thread. Catalog and co-marketing are now blocked on authenticated portal work on our side; the Pavilion is invite-only. The only reply to send is the narrow two-question draft for Startup Showcase routing and an ACE contact. The old 2026-09-25 no-response follow-up is cancelled.
Still ours to finish regardless of the reply: the portal record correction (step 1
of nvidia-apps-catalog-listing.md), because a
curator who opens the record after reading the email should not find a CUDA
consumer with DeepVariant checked, and the second product record for the
<agent-3d> embed, which targets the Digital Humans workload filter the main
record does not.
Verification log
| Claim | Checked how | Result |
|---|---|---|
| GTC 2026 CFP closed; 2027 is next | Fetched call-for-submissions | "now closed" for GTC 2026; NVIDIA's FAQ page already titled GTC San Jose 2027 |
| GTC 2027 CFP still shut (2026-09-02) | Re-fetched the same page, plus a search across event trackers | Same "now closed" GTC 2026 copy; no 2027 CFP or registration open; trackers report 2027-03-15 to 2027-03-18 San Jose |
| A2F lane live and its latency (2026-09-02) | Three POSTs each to https://three.ws/api/a2f, audio-in and text-in |
200 on all six; 4.64 s of speech animates in 1.29 to 2.26 s, text-to-speech-to-animation in 1.88 to 3.41 s; 140 frames at 30 fps, 55 blendshapes |
| Inception benefit wording | Fetched nvidia.com/en-us/startups | Quoted verbatim above |
| Social handles | Fetched NVIDIA social directory | Full list; the table above is the topic-relevant subset |
| AI Podcast guest form | Fetched ai-podcast.nvidia.com | Embedded "Submit Guest Ideas" form, fields recorded above |
| NGC publishing prerequisites | Searched the NGC Software Partner page | Legal agreement, staging repo, security scan, QA, sign-off |
| ISV program retired | WebFetch on nvidia.com/en-us/programs/isv/ |
HTTP 301 to developer.nvidia.com |
| All four forum posts live | curl -o /dev/null -w '%{http_code}' |
All 200 |
/nvidia never announced on X |
Read announcement-coverage.md inventory | Column empty for /nvidia, populated for /openai |
| Jensen Huang engagement has no source | data/timeline.json, event 2026-07-18-jensen-huang |
source_url: null, as are 4 other NVIDIA events |
| GPU fleet shape | gcloud run services list across all regions, filtered to accelerator node selectors |
8 workers, 12 deployments, 11 L4 + 1 RTX PRO 6000 |
| GPU fleet shape, re-counted 2026-09-02 | Same command, isolated CLOUDSDK_CONFIG |
Unchanged: 8 services, 12 deployments, 11 L4 + 1 RTX PRO 6000. model-trellis is live in us-central1 and us-east4. |
Recount the fleet before quoting it anywhere:
gcloud run services list --project aerial-vehicle-466722-p5 \
--format="csv[no-heading](metadata.name,region,spec.template.spec.nodeSelector)" \
| grep -i accelerator
Related
- Big-tech recognition dispatch board: this map ranked against every other big-tech surface (OpenAI, Anthropic, AWS, IBM) by the odds of a published mention this week. Read that first if you are deciding where a week goes, and this one once NVIDIA is the answer.
- IBM visibility map: the same surface-by-surface treatment for IBM
- NVIDIA Inception membership: what membership is, and the rule that it is a program and not a partnership or an endorsement
- Apps Catalog listing kit and inclusion request: the curated marketing listing
- NGC Catalog listing kit: the self-serve software listing, its prerequisite audit, and the container that clears it
- NVIDIA models on three.ws: the source of truth for every NVIDIA technical claim made in any pitch
- Listings and distribution: the canonical program and directory inventory
- Announcement coverage: which surfaces have been announced and which have not
- Partners: the independence and no-endorsement language that must survive any rewrite