Available Models

Pass the exact model id string (left column) together with a task_type the model supports. An unrecognized model or task_type, or a pair the model does not support, returns HTTP 400.

Hosting is either MachGen (we run the model on our own GPUs) or Partner (forwarded to a third-party provider). It does not change how you call the API.

Video models

Task alternatives:

• Kling-v3 does not support R2V; use Kling-o3 for that workflow.

End frame: the model's I2V also accepts an optional second image used as the last frame (src_image_urls entry 1).

Upscale: task_type=UPSCALE on a completed video; MiniMax-H3 accepts only its own 768p outputs and returns 2K.

Image models

Upscale: task_type=UPSCALE on a completed image; the output size is derived from the source and upscale_config.factor.

Audio models

T2S speech - T2D multi-speaker dialogue - T2SFX sound effects - T2M music.

Seedance-2.5 draft mode makes a 480p sample, then a separate 1080p final. See Seedance 2.5.

YuE2 music

For accounts with access to YuE2-3B, use task_type="T2M", a musical style in prompt, and 400 to 10,000 characters in audio_config.lyrics for vocal songs. For instrumental tracks, set audio_config.force_instrumental=true. In Generate, turn on Instrumental only under Advanced. Optional lyrics or section tags in audio_config.lyrics help shape the track's flow and duration. Lyrics are used only to plan the music. For example, use [Intro], [Verse], [Chorus], and [Outro] to suggest a structure. Tempo and section tags guide the arrangement, but exact duration and section order are not guaranteed. Instrumental lyrics have no minimum length and a maximum of 10,000 characters.

Instrumental mode does not guarantee the absence of vocals. Generated tracks may still include singing or wordless vocalizations, such as humming. Use prompt to describe the genre, instruments, mood, and tempo; place any lyrics in audio_config.lyrics.

YuE2 returns FLAC audio capped at approximately 6 minutes. Tracks that reach this limit are truncated and returned successfully, with the ending cut off. Shorter lyrics help the song finish within the limit.

Score generation can also reach its limit and fail without audio. See the score-limit error and recovery guidance.