Text-to-video and image-to-video from China's top video models through one OpenAI-compatible async API. Per-second or per-token billing, failures auto-refunded, and a genuinely free tier.
Nine video models across four providers, one endpoint. Draft for free on CogVideoX Flash, then scale to Seedance 2.5 (with synchronized native audio) or Hailuo H3 for cinematic output.
| Model | Price | Strength | |
|---|---|---|---|
| doubao-seedance-2.5 | 480p $0.14/s · 720p $0.32/s | Native synchronized audio, up to 15s clips | |
| doubao-seedance-2.0 | from $0.067/s (480p) to $1.00/s (4K) | Up to 4K resolution | |
| doubao-seedance-2.0-fast | from $0.054/s (480p) | Cheapest 4K-capable video | |
| minimax-video-h3 | ~$0.077/s ($1.54/1M tok) | Hailuo flagship T2V, cinematic motion | NEW |
| minimax-video-h3-max | ~$0.077/s ($1.54/1M tok) | Hailuo top-quality T2V | NEW |
| minimax-video-v2.3 | ~$0.062/s ($1.54/1M tok) | Hailuo text-to-video | NEW |
| minimax-video-v2.3-fast | ~$0.042/s ($1.54/1M tok) | Hailuo image-to-video, lowest cost | NEW |
| hy-video-v1.5 | ~$0.046/s ($1.54/1M tok) | Tencent Hunyuan T2V + I2V | NEW |
| cogvideox-flash | FREE | Zhipu CogVideoX, free unlimited tier | FREE |
/v1/video/generations) for every model - submit, poll, get the mp4 URLcurl -X POST https://aiapi-pro.com/v1/video/generations \
-H "Authorization: Bearer $NOVAI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "minimax-video-h3",
"prompt": "A red sports car on a coastal road at sunset, cinematic",
"resolution": "720p",
"duration": 5,
"ratio": "16:9"
}'
# -> { "id": "", "status": "queued" } curl "https://aiapi-pro.com/v1/video/generations/{task_id}?model=minimax-video-h3" \
-H "Authorization: Bearer $NOVAI_API_KEY"
# -> { "status": "processing" } ... then when done:
# -> { "status": "succeeded", "video_url": "https://.../out.mp4" }POST /v1/video/generations returns {id, status:"queued"}. Poll GET /v1/video/generations/{id} until status is succeeded (then video_url holds the mp4) or failed (auto-refunded). An estimate is pre-deducted at submit and reconciled to the real token count on success. Seedance models bill per output second instead.Start free with cogvideox-flash for drafts. Use doubao-seedance-2.5 when you need synchronized native audio or up-to-15s clips, minimax-video-h3 for cinematic fidelity, and doubao-seedance-2.0-fast for the cheapest 4K output.
Seedance models bill per output second (resolution-tiered). Hailuo and Hunyuan bill per token at $1.54/1M (~$0.042-0.077/s). An estimate is pre-deducted at submit and reconciled to real usage on success; failures are refunded in full.
Video uses an async task API (submit + poll) at /v1/video/generations rather than the synchronous OpenAI chat shape, but it is called with the same API key and base URL as everything else on NovAI.
Chinese providers may burn a visible AI-generated label into video output under China's labelling rules; NovAI cannot disable it. The free CogVideoX tier also carries a NovAI watermark until you make any top-up. See /watermark-notice.
Deep-dive articles, benchmarks and real tests for these video models — including pricing vs the leading alternatives.