# Replicate video models (slim)

> Replicate runs third-party and open video models, among them Veo 3.1, Kling 3.0, Runway Gen-4.5, Seedance 2.0 and Wan, through one predictions API with Python and JavaScript clients and an MCP server.

- Full: https://www.anchorterminal.com/tools/replicate-video.md (~8,100 tokens) · this version ~1,680 tokens · JSON https://www.anchorterminal.com/tools/replicate-video.json · canonical https://www.anchorterminal.com/tools/replicate-video
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · Updated: 2026-10-09

**B · 62.1/100 · rank #402 of 842 · #4 in Video generation · not agent-ready · confidence medium**


Assessment: One bearer token reaches more than twenty official video models at published per-second prices, each with a typed input schema, and failed runs aren't charged. Tokens have no scopes or spend cap, no SLA is published, and the terms let Replicate withdraw any model without notice.

## Facts

- Kind: Model platform · vendor: Replicate (Cloudflare) · category: Video generation · legal entity: Replicate, LLC · provenance 87/100
- Endpoint: `https://api.replicate.com/v1` (HTTP, SSE (legacy), stdio)
- Auth: API key · pricing: Pay per use · x402: no · licence: Proprietary service under Replicate's terms of service. The Python and JavaScript clients and `replicate-mcp` are Apache-2.0, and each model carries its own licence
- Probe metrics: not measured yet (probes haven't run)
- Models: Veo 3.1, 3.1 Fast and 3.1 Lite, Kling 3.0 and 3.0 Omni, Runway Gen-4.5, Seedance 2.0 and 2.0 Fast, Wan 2.5, 2.7 and 3.0, Hailuo 2.3, Grok Imagine Video, Luma Ray 3.2, Vidu Q3, PixVerse v6, Pruna p-video, plus community models
- Max clip length: Set by each model. Veo 3.1 4, 6 or 8 seconds, Gen-4.5 5 or 10, Kling 3.0 3 to 15, Wan 3.0 2 to 30
- Resolution: Set by each model. Veo 3.1 720p or 1080p, Wan 3.0 480p to 1080p, Kling 3.0 and Seedance 2.0 up to 4K
- Audio: Native audio on Veo 3.1, Kling 3.0, Seedance 2.0, Grok Imagine Video, Vidu Q3 Pro, Wan 2.5 and p-video per the collection page. Veo 3.1 and Kling 3.0 charge more with audio on
- Typical job time: Async prediction, polled or sent to a webhook. No per-model times published. The collection page says Grok Imagine Video takes about 30 seconds
- Free tier: None standing. A new account can run `minimax/video-01` a limited number of times from the try-for-free collection before adding billing
- Rate limits: 600 create-prediction requests a minute, 3,000 a minute for other endpoints, 6 a minute on granted credit with no card
- Output licence: Set by each model's own terms and Replicate's terms of service. The collection page says most models allow commercial use and tells users to check the model page
- Prices: Veo 3.1 with audio $0.40 per second of video; Veo 3.1 Fast with audio $0.15 per second of video; Kling 3.0 pro (1080p) $0.224 per second of video; Runway Gen-4.5 $0.12 per second of video; Seedance 2.0 720p $0.18 per second of video; Wan 2.7 text-to-video $0.10 per second of video; Wan 3.0 1080p $0.10 per second of video; Hailuo 2.3 768p $0.0467 per second of video

- Scores: Reliability 75, Performance pending, Schema & documentation 86, Agent ergonomics 72, Security & auth 41, Payments & pricing 30, Task success pending, Maintenance & community 48, Transparency & trust 72 · total over the 7 assessed categories
- Why: Reliability, Graded as a hosted service, on the predictions API. · Schema & documentation, Public OpenAPI document with 26 paths and 36 operations, and each model publishes its own input and output schema (25). · Agent ergonomics, A prediction returns a small JSON object with the output as a file URL. · Security & auth, Bearer tokens with the prefix `r8_`, several per account, each named and able to be disabled. · Payments & pricing, No x402, MPP or L402 (0). · Maintenance & community, The brief counts the last published model for a closed service. · Transparency & trust, Closed platform under terms of service dated 1 April 2026 that name Replicate, LLC.
- Sources: 40, open questions: 7, both in the full twin
- Capabilities: video.generate, video.image-to-video, video.audio, video.edit, video.extend
- JSON: https://www.anchorterminal.com/api/v1/tools/replicate-video.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/replicate-video.svg` or a link to https://www.anchorterminal.com/tools/replicate-video from a page on replicate.com or one of its subdomains, or the README of github.com/replicate/replicate-python, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. Call official models at `POST /v1/models/{owner}/{name}/predictions` with no version, then poll `GET /v1/predictions/{id}` or pass `webhook`. Video jobs usually outlast the 60-second `Prefer: wait` window
2. Download the output file within the hour. API predictions lose inputs, outputs and logs after 60 minutes
3. Read the model's price tiers before setting `resolution`, `mode` or `generate_audio`. Kling 3.0 runs from $0.168 to $0.42 a second and Veo 3.1 doubles with audio
4. Send `Cancel-After` on every create. There is no idempotency key, so a blind retry is a second paid video
5. Keep credit above $20 or enable auto-reload. Low balances are throttled, and granted credit with no card is held to 6 requests a minute

## Connect

```bash
curl -X POST https://api.replicate.com/v1/models/google/veo-3.1/predictions \
  -H "Authorization: Bearer $REPLICATE_API_TOKEN" -H "Content-Type: application/json" \
  -d '{"input":{"prompt":"a red fox running through fresh snow","duration":4,"resolution":"720p"}}'
```

```bash
claude mcp add replicate https://mcp.replicate.com/sse --transport sse --scope user
```

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| Runway API | C | 60.8 | video.generate, video.image-to-video, video.edit, video.audio, video.extend | https://www.anchorterminal.com/tools/runway.min.md |
| Pika API | D | 47.3 | video.generate, video.image-to-video, video.edit, video.audio, video.extend | https://www.anchorterminal.com/tools/pika.min.md |
| PixVerse API | E | 42.4 | video.generate, video.image-to-video, video.edit, video.audio, video.extend | https://www.anchorterminal.com/tools/pixverse.min.md |
| Kling AI API | F | 21.7 | video.generate, video.image-to-video, video.edit, video.audio, video.extend | https://www.anchorterminal.com/tools/kling.min.md |
| fal video models | B | 68.9 | video.generate, video.image-to-video, video.edit, video.audio | https://www.anchorterminal.com/tools/fal-video.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)