API model string
not applicable (no API exists)Context window
n/a
Max output n/a
- Knowledge cutoff
- n/a
- Released
- March 2025 (v1.1 checkpoint June 2025)
- Modalities
- image / multi-view image in, 3D video out
Last verified July 10, 2026
Rate limits by tier
Stable Virtual Camera is not listed on platform.stability.ai — there is no API endpoint, no credit cost, and no rate limits to document. The two real access paths are the Hugging Face Space demo (non-commercial, queue-based) and self-hosted open weights under a non-commercial research license.
| Tier | Requirements | RPM | TPM | RPD | Concurrent | Notes |
|---|---|---|---|---|---|---|
| Stability Developer Platform API | n/a | n/a | n/a | n/a | n/a | Stable Virtual Camera is NOT listed on platform.stability.ai. No API endpoint, no Stability credits cost, no rate limits to document here. Stability credits (1 credit = $0.01) apply only to listed services such as Stable Image Ultra, Stable Image Core, and the SD3.5 family. |
| Hugging Face Space (free interactive demo) | Hugging Face account | not published (shared compute, queue-based) | n/a | not published | not published (queue-based shared GPU) | Interactive demo only — not a production API. Non-commercial use only. Queue times vary depending on shared compute load. Not suitable for automation or production workloads. |
| Self-host (open weights) | Own GPU hardware; acceptance of Stability AI Non-Commercial Research Community License | self-managed | self-managed | self-managed | self-managed | Non-commercial use only — commercial use is NOT permitted under the Stability AI Non-Commercial Research Community License. This is distinct from the Stability AI Community License (which covers SD3.5/SDXL for commercial use under $1M revenue). Weights: huggingface.co/stabilityai/stable-virtual-camera. |
Swipe the table sideways to see every limit column.
- 1.Stability credits (1 credit = $0.01) apply only to listed API services (Stable Image Ultra, Core, SD3.5 family) — they do NOT apply to Stable Virtual Camera.
- 2.New Stability API accounts receive 25 free credits one-time (Google signup only) — these apply to listed API services, not Stable Virtual Camera.
- 3.Stable Video (a different Stability model) was also removed from the Stability API and is now self-host license only.
- 4.SD1.6 API was deprecated July 24, 2025 — separate from Stable Virtual Camera.
- 5.No pricing exists for Stable Virtual Camera — there is no API.
- 6.For reference only, Stability API services that DO have credit-based pricing: Stable Image Ultra $0.08/image (0.8 credits); Stable Image Core $0.03/image; SD3.5 family pricing varies. These are entirely separate services from Stable Virtual Camera.
- 7.Self-host cost = GPU infrastructure only. Model is approximately 1.3B parameters; runs on a single A100 40GB or two A6000 48GB GPUs for research workloads.
Limits verified against the Stability AI docs, July 10, 2026.
not applicable (no API exists) vs the alternatives
No direct API competitor exists for novel-view-synthesis and 3D camera control via a per-token or per-image paid API as of July 2026. Comparison is against live API alternatives for teams needing a production-ready solution.
| Aspect | not applicable (no API exists) | Gemma 3 | Llama 4 Scout |
|---|---|---|---|
| API availability | none (no hosted API) | live Google API (Gemma 3) | third-party hosted (Llama 4 Scout) |
| Commercial use | not permitted (non-commercial research license) | yes (Gemma License, commercial) | yes (Llama 4 Community License) |
| License permissiveness | Non-Commercial Research Community License (restrictive) | Gemma License (commercial up to terms) | Llama 4 Community License |
| Cost per inference | $0 (self-host, non-commercial research) | $0.040/MTok input (Gemma 3) | $0.08-$0.11/MTok input (Llama 4 Scout) |
| Production readiness | research preview only | GA | GA |
| Task fit (image to 3D video / novel view synthesis) | purpose-built for novel-view-synthesis and 3D camera control | not applicable (text/vision LLM) | not applicable (text/vision LLM) |
Swipe the table sideways to see every model.
Hitting a 429? The playbook
The exact errors you'll see
Not applicable — no Stable Virtual Camera API endpoint exists.If hitting Stability API credit-limit errors on OTHER listed Stability services: HTTP 429 or credit-exhaustion response from platform.stability.aiStagingAreaFull or QueueFull (Hugging Face Space demo — internal queue; wait and retry)Why it happens & how to fix it
Searching for a Stable Virtual Camera REST API endpoint
The endpoint does not exist. Use the Hugging Face Space demo (huggingface.co/spaces/stabilityai/stable-virtual-camera) for interactive testing, or self-host the open weights under the non-commercial research license.
Stability API credits exhausted on other Stability services (not Stable Virtual Camera)
Add credits at platform.stability.ai. Note: these credits apply to Stable Image Ultra, Core, SD3.5 — not to Stable Virtual Camera.
Self-hosted inference running out of GPU memory (OOM)
Reduce batch size to 1; switch from float32 to float16 precision; the model requires approximately 8GB VRAM minimum for small checkpoints. Use the v1.1 checkpoint (June 2025) which includes quality improvements.
Retry strategy
No API retry logic is applicable for Stable Virtual Camera. For the Hugging Face Space demo, simply wait for the queue and resubmit. For Stability API credit-based services (unrelated to SVC), standard HTTP 429 backoff applies: honor Retry-After if present, otherwise exponential backoff with jitter.
1// Stable Virtual Camera has no REST API endpoint.2// To run locally via the Hugging Face model card:3// pip install git+https://github.com/Stability-AI/stable-virtual-camera4// Accept the non-commercial license at huggingface.co/stabilityai/stable-virtual-camera5// Follow the README for inference examples.6//7// For production use requiring an API, consider Stability AI's listed services8// (Stable Image Ultra, SD3.5 family) or contact stability.ai/contact9// regarding commercial licensing for Stable Virtual Camera.1011// Example: Stability API client for LISTED services (not SVC)12import fetch from 'node-fetch';1314async function generateStableImage(prompt: string): Promise<Buffer> {15 const response = await fetch(16 'https://api.stability.ai/v2beta/stable-image/generate/ultra',17 {18 method: 'POST',19 headers: {20 Authorization: `Bearer ${process.env.STABILITY_API_KEY}`,21 Accept: 'image/*',22 },23 body: JSON.stringify({ prompt, output_format: 'webp' }),24 }25 );26 if (response.status === 429) {27 // Credit exhaustion or rate limit — add backoff and retry28 throw new Error('Stability API credits exhausted or rate limit hit');29 }30 if (!response.ok) throw new Error(`Stability API error: ${response.status}`);31 return Buffer.from(await response.arrayBuffer());32}How to raise your limits
The ladder from the starter tier to enterprise — what each rung takes, and what it unlocks.
Hugging Face Space demo
ImmediateVisit huggingface.co/spaces/stabilityai/stable-virtual-camera — no setup required
Unlocks: Free interactive demo for evaluation and prototyping; non-commercial use only; queue-dependent; not suitable for production or automation
Self-host (non-commercial research)
GPU infrastructure setup time (hours to days)Accept the non-commercial license at huggingface.co/stabilityai/stable-virtual-camera; follow install instructions in the GitHub repo (github.com/Stability-AI/stable-virtual-camera)
Unlocks: Non-commercial research use at your own scale; full control over throughput; single A100 40GB or two A6000 48GB GPUs recommended; use v1.1 checkpoint for best quality
Commercial licensing
Unknown (licensing negotiation required; no self-serve commercial path existed as of July 2026)Contact Stability AI directly at stability.ai/contact to inquire about commercial licensing terms
Unlocks: Potential commercial use rights — not available via self-serve or any third-party API host as of July 2026
Cut your token spend
Use the Hugging Face Space for evaluation and prototyping
Zero cost, zero setup — free shared compute for non-commercial testingVisit huggingface.co/spaces/stabilityai/stable-virtual-camera. Queue times vary by load. This is the fastest path to evaluate whether Stable Virtual Camera fits your research use case before committing to a self-hosted setup.
Self-host with the v1.1 checkpoint for best quality
Quality improvement vs launch weights (per Hugging Face model card)Download the June 2025 v1.1 checkpoint from huggingface.co/stabilityai/stable-virtual-camera — it is newer than the March 2025 launch weights and includes quality improvements documented on the model card.
Run on a single A100 40GB or two A6000 48GB GPUs
Manageable hardware cost for research workloads (~1.3B parameters)The model is approximately 1.3B parameters — significantly smaller than text LLMs. A single A100 40GB is sufficient for most research inference runs. Use float16 precision to minimize VRAM usage.
Monitor Stability AI's API roadmap
First to know if commercial API access becomes availableWatch stability.ai/news-updates and the Stability AI Discord. Stable Video was previously on the API and then removed; the reverse is possible for Stable Virtual Camera if demand is sufficient.
Evaluate third-party video AI APIs for commercial novel-view-synthesis needs
Unblocks commercial production while awaiting official Stable Virtual Camera licensingFor commercial-grade novel-view-synthesis or 3D video generation, evaluate Runway, Pika, or Kling APIs as interim solutions. No direct Stability API equivalent exists for this task category today.
Frequently asked questions
Does Stable Virtual Camera have an API?
No. As of July 2026, Stable Virtual Camera is not listed on the Stability AI Developer Platform (platform.stability.ai) and has no hosted API endpoint. There is no REST API, no SDK integration, and no Stability credits pricing for this model. Access is limited to the Hugging Face Space demo (non-commercial, queue-based) and self-hosted open weights under a non-commercial research license.
How do I use Stable Virtual Camera?
You have two paths: (1) Interactive demo — visit huggingface.co/spaces/stabilityai/stable-virtual-camera for free non-commercial evaluation with no setup required. (2) Self-host — accept the non-commercial license at huggingface.co/stabilityai/stable-virtual-camera and follow the GitHub repo setup guide (github.com/Stability-AI/stable-virtual-camera). Self-hosting requires a GPU with at least 8GB VRAM; an A100 40GB is recommended for smooth inference.
Can I use Stable Virtual Camera commercially?
No — commercial use is not permitted under the Stability AI Non-Commercial Research Community License. This is different from the Stability AI Community License (which covers Stable Diffusion 3.5 and SDXL for commercial use below $1M annual revenue). If you need commercial rights to Stable Virtual Camera, contact Stability AI directly at stability.ai/contact — no self-serve commercial licensing option was available as of July 2026.
What is Stable Virtual Camera used for?
Stable Virtual Camera is a purpose-built model for novel-view-synthesis: generating 3D video sequences and new camera perspectives from one or more input images. It is a research model designed to explore camera control and 3D scene understanding, not a general-purpose image or text generation model.
Is Stable Virtual Camera free?
The Hugging Face Space demo is free (shared compute, queue-based). Self-hosting the open weights is free in terms of per-inference cost — you pay only for your own GPU infrastructure. However, both are restricted to non-commercial research use. There is no paid commercial API option.
How is Stable Virtual Camera different from Stable Video Diffusion?
They are separate models. Stable Video Diffusion generates videos from images but focuses on temporal coherence. Stable Virtual Camera specifically targets novel-view-synthesis and 3D camera control — generating what a scene looks like from a different camera angle. Both are available as open weights only (no production API); Stable Video was also removed from the Stability API.
What API should I use if I need novel-view-synthesis commercially?
No direct commercial API exists for this task category as of July 2026. For commercial-grade 3D video generation, evaluate third-party video AI APIs such as Runway, Pika, or Kling as interim alternatives. RapidDev (rapidevelopers.com/contact) can help evaluate your specific use case and identify the best available tooling for production novel-view-synthesis workflows.
We build AI apps that don't hit rate limits
- Retry, backoff & caching built in
- Multi-provider fallback routing
- Fixed price, you own the code
30-min call. No commitment.