Model Grouping
Nano Banana
Primarily used for image generation and editing; also supports Gemini native format integration. | Model | Primary Integration Method | |---|---| |nano-banana | /v1/images/generations / /v1/images/edits / /v1beta/... |
| nano-banana-2 | /v1/images/generations / /v1/images/edits / /v1beta/... |
| nano-banana-2-2k | /v1/images/generations / /v1/images/edits / /v1beta/... |
| nano-banana-2-4k | /v1/images/generations / /v1/images/edits / /v1beta/... |
| nano-banana-pro | /v1/images/generations / /v1/images/edits / /v1beta/... |
| nano-banana-pro-2k | /v1/images/generations / /v1/images/edits / /v1beta/... |
| nano-banana-pro-4k | /v1/images/generations / /v1/images/edits / /v1beta/... |
Sora
Prefer the/v1/videos OpenAI video-compatible format, which is suitable for unified video integration.
| Model | Primary Integration Method |
|---|---|
| sora-2 | /v1/videos |
| sora-2-pro | /v1/videos |
Gemini Omni
gemini-omni uses new-api to expose public model names, which can be invoked via /v1/videos or /api/v3/contents/generations/tasks; use mode=t2v/r2v/edit to distinguish text-to-video, reference-based generation, and video editing, and the duration will be automatically mapped to the 4 / 6 / 8 / 10-second tiers.
| Model | Primary Integration Method |
|---|---|
| gemini-omni | /v1/videos / /api/v3/contents/generations/tasks |
HappyHorse
Supports Alibaba DashScope official video generation format, and also supports compatible integration via/v1/videos.
| Model | Primary Integration Method |
|---|---|
| happyhorse-1.0 | /api/v1/services/aigc/video-generation/video-synthesis / /v1/videos |
| happyhorse-1.0-i2v | /api/v1/services/aigc/video-generation/video-synthesis / /v1/videos |
| happyhorse-1.0-t2v | /api/v1/services/aigc/video-generation/video-synthesis / /v1/videos |
| happyhorse-1.0-r2v | /api/v1/services/aigc/video-generation/video-synthesis / /v1/videos |
| happyhorse-1.0-video-edit | /api/v1/services/aigc/video-generation/video-synthesis / /v1/videos |
Wan
Refer to the specific API documentation. | Model | Primary Integration Method | |---|---| |wan-3.0 | /v1/videos |
| wan-3.0-480p | /v1/videos |
| wan-3.0-1080p | /v1/videos |
Veo
For the Veo primary model, prefer the/v1/videos compatible format; /v1beta/... reserves notes for the Veo native protocol. gemini-omni does not use the Google native path; please use /v1/videos or the Seedance task format.
| Model | Primary Integration Method |
|---|---|
| veo3 | /v1/videos / /v1beta/... |
| veo3-fast | /v1/videos / /v1beta/... |
| veo3-fast-frames | /v1/videos / /v1beta/... |
| veo3-frames | /v1/videos / /v1beta/... |
| veo3-pro | /v1/videos / /v1beta/... |
| veo3-pro-frames | /v1/videos / /v1beta/... |
| veo3.1 | /v1/videos / /v1beta/... |
| veo3.1-4k | /v1/videos / /v1beta/... |
| veo3.1-components | /v1/videos / /v1beta/... |
| veo3.1-fast | /v1/videos / /v1beta/... |
| veo3.1-pro | /v1/videos / /v1beta/... |
| veo3.1-pro-4k | /v1/videos / /v1beta/... |
Grok
The currently released model is a video generation model; it is recommended to prioritize integration via the unified video interface. | Model | Primary Integration Method | |---|---| |grok-imagine-1.0-video | /v1/videos |
| grok-imagine-video-1.5-preview | /v1/videos |
| grok-video-3 | /v1/videos |
Seedance
Supports the/api/v3/contents/generations/tasks native format, and also supports compatible integration via /v1/videos; gemini-omni can reuse this task format and use mode=t2v/r2v/edit to differentiate capabilities.
| Model | Primary Integration Method |
|---|---|
| doubao-seedance-1-0-lite-i2v-250428 | /api/v3/contents/generations/tasks / /v1/videos |
| doubao-seedance-1-0-lite-t2v-250428 | /api/v3/contents/generations/tasks / /v1/videos |
| doubao-seedance-1-0-pro-250528 | /api/v3/contents/generations/tasks / /v1/videos |
| doubao-seedance-1-0-pro-fast-251015 | /api/v3/contents/generations/tasks / /v1/videos |
| doubao-seedance-1-5-pro-251215 | /api/v3/contents/generations/tasks / /v1/videos |
| doubao-seedance-2-0-260128 | /api/v3/contents/generations/tasks / /v1/videos |
| doubao-seedance-2-0-fast-260128 | /api/v3/contents/generations/tasks / /v1/videos |
Kling Images
Provides Kling image generation, Omni images, and virtual try-on capabilities externally via/v1/images/generations.
| Model | Primary Integration Method |
|---|---|
| kling-image | /v1/images/generations |
| kling-omni-image | /v1/images/generations |
| kolors-virtual-try-on-v1 | /v1/images/generations |
| kolors-virtual-try-on-v1-5 | /v1/images/generations |
Kling Videos
Kling 4.0 directly supports the standard OpenAI/Sora parameters prompt/seconds/size/input_reference, as well as Seedance parameters content/duration/resolution/ratio/mode, with no additional contents/settings/options required. The legacy envelope remains compatible. mode=t2v/i2v/i2v_first_last/reference_images/reference_material distinguishes scenarios; when reference materials are provided, the default is reference_material, and when none are provided, the default is t2v. | Model | Primary Integration Method | |---|---| |kling-4.0 | /api/v3/contents/generations/tasks / /v1/videos / /text-to-video/kling-4.0 / /image-to-video/kling-4.0 / /omni-video/kling-4.0 |
| kling-4.0-1080p | /api/v3/contents/generations/tasks / /v1/videos / /text-to-video/kling-4.0 / /image-to-video/kling-4.0 / /omni-video/kling-4.0 |
| kling-video | /api/v3/contents/generations/tasks / /v1/videos |
| kling-* | /api/v3/contents/generations/tasks / /v1/videos |
MiniMax / Hailuo
Use the unified video capability for integration.minimax-h3 Supports text-to-video, image references, and audio references; hailuo-2.3 requires a first-frame image and is suitable for first-frame image-to-video.
| Model | Primary Integration Method |
|---|---|
| minimax-h3 | /v1/videos / /api/v3/contents/generations/tasks |
| hailuo-2.3 | /v1/videos / /api/v3/contents/generations/tasks |
| hailuo* | /v1/videos / /api/v3/contents/generations/tasks |
Vidu
Vidu Q4:vidu-q4 is fixed at 720p, and vidu-q4-540p is fixed at 540p. Supports generating 3–16 second videos. At least 1 image is required, with up to 15 reference images supported. Reference-based generation can include up to 3 MP3 audio clips, each 3–12 seconds and up to 50 MB. Pure text-to-video, specified first and last frames, and reference videos are not supported. Multiple images are used as shared references and do not indicate first/last-frame positions. Tasks can be created through POST /v1/videos, POST /api/v3/contents/generations/tasks, or the Vidu formats POST /ent/v2/img2video and POST /ent/v2/reference2video. For the unified entry point, mode=reference_images is recommended; use mode=reference_material when including audio. Reference audio can use audio_urls; the Seedance format can also use the audio_url field of the standard content. The audio boolean field controls whether audio is generated and is used separately from the reference audio URL list. The resolution must match the model name. Legacy Q3/Q2/V2 models should continue using their original APIs according to their respective capabilities.
| Model | Primary Integration Method |
|---|---|
| vidu-q4 | /ent/v2/... / /v1/videos |
| vidu-q4-540p | /ent/v2/... / /v1/videos |
| viduq3-turbo | /ent/v2/... / /v1/videos |
| viduq2 | /ent/v2/... / /v1/videos |
| vidu2.0 | /ent/v2/... / /v1/videos |
| vidu* | /ent/v2/... / /v1/videos |
For pricing information, visit: Pricing page.