New video generation models on BazaarLink: Wan 3.0, Veo 3.1, MiniMax H3, Grok Imagine Video, Seedance 2.x, Kling
Six new model families join the video API, on the same /api/v1/videos endpoint and the same API key.
The BazaarLink video API now includes the models below. They all use the same asynchronous interface: submit a job, receive a job id, then poll for the result. Existing video integrations keep working; to try a new model, change the model field.
What's new
| Family | model value | Resolutions you can request |
|---|---|---|
| Wan 3.0 | alibaba/wan3.0 | 480p / 720p / 1080p |
| Veo 3.1 | google/veo-3.1 | 720p / 1080p / 4K (audio optional) |
| MiniMax H3 | minimax/minimax-h3 | 720p / 1080p |
| Grok Imagine Video | x-ai/grok-imagine-video | 480p / 720p |
| Grok Imagine Video 1.5 | x-ai/grok-imagine-video-1.5 | 480p / 720p |
| Seedance 2.0 Mini | bytedance/seedance-2.0-mini | 480p / 720p |
| Seedance 2.0 | bytedance/seedance-2.0 | 480p / 720p / 1080p |
| Seedance 2.5 | bytedance/seedance-2.5 | 480p / 720p / 1080p |
The Kling family is also available (text-to-video and image-to-video, standard and pro tiers):
klingai/kling-v2.5-turbo-t2v-std,klingai/kling-v2.5-turbo-i2v-stdklingai/kling-v2.6-t2v,klingai/kling-v2.6-t2v-std,klingai/kling-v2.6-i2vklingai/kling-v3.0-t2v,klingai/kling-v3.0-t2v-std,klingai/kling-v3.0-i2v-std
Kling has no resolution parameter; the standard and pro tiers are separate model names.
How to call it
curl https://api.bazaarlink.ai/api/v1/videos \
-H "Authorization: Bearer $BAZAARLINK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "alibaba/wan3.0",
"prompt": "A small red bird flying over a calm blue lake",
"duration": 4,
"resolution": "720p"
}'
The response contains a job id. Poll the job by id, and when it completes download the video from the link in the result. Video links point to BazaarLink.
Things to know
- Set
resolutionexplicitly. When no resolution is given, Wan 3.0 outputs 1080p by default; we recommend always specifying the resolution you want. - Parameters differ per model. A resolution the model does not support is rejected up front; see each model page for the exact options.
- Billed on the delivered length. Pricing is per second, by resolution (and for some models by whether audio is generated), shown on each model page; the charge is based on the video actually delivered.
- If a parameter combination does not behave as documented, send us the job id and we will look into it.
TWD billing · Taiwan invoices · leading AI models · OpenAI-compatible API