MiniMax H3 Max (BudgetPixel)

MiniMax H3 Max — the post-trained H3 variant tuned for stronger prompt adherence and aesthetics, with native audio. Text-to-video; image-to-video (start and/or end frame); reference-to-video mixing up to 5 reference images, up to 3 reference video clips (2-15s each, 15s combined — billed per second of INPUT, see POST /v1/cost) and up to 3 reference audio clips (free); and video extension (supply a `video` to continue it). 480p/768p/1080p (2K when extending), any duration 5-15 seconds. Priced per output second by resolution (see `resolution_pricing` in GET /v1/models); reference images beyond the first two add a small per-image charge. Pricing: 105 credits per second.

By budgetpixel-ai·Created 25 days ago·Updated about 14 hours ago· 4

audio
end_image
image
reference_audios
reference_images
reference_videos
video

video

◄prompt►

◄aspect_ratio16:9►

◄length_seconds5►

◄resolution768p►

◄seed-1►

◄force_rerunfalse►

CategoryBudgetPixel/Video

Inputs (13)

Name	Type	Default	Description
prompt	STRING		Text description of the video. In reference mode, address assets by order: "Image 1", "Video 1", "Audio 1". When extending, describe what happens next, not the source.
aspect_ratioopt	COMBO	16:9	Aspect ratio. Text-to-video uses it directly; image-to-video follows the frame image and ignores it; reference-to-video also accepts "adaptive" (the default there); extension accepts "auto" (follow the source — the default there) or a ratio to crop to. Defaults: 16:9 for text-to-video.
audioopt	AUDIO		Optional soundtrack (≥2s, ≤15MB, MP3/WAV) pinned to a text-to-video or image-to-video output: it replaces the generated audio, trimmed to the video length. Provide a public audio URL or an uploaded-file URL from POST /v1/uploads. Not available with reference media or `video`.
end_imageopt	IMAGE		Optional last frame. With `image` the video interpolates between the two frames; on its own it is an end-keyframe generation. Same input forms as `image`.
imageopt	IMAGE		Optional first frame for image-to-video — the output follows its framing. Provide a public image URL, a data URI, raw base64, or an uploaded-file URL from POST /v1/uploads. Can't be combined with reference media or `video`.
length_secondsopt	INT	55–15	Output length in seconds (5-15). When extending, the seconds of NEW footage appended to the source.
reference_audiosopt	AUDIO		Optional reference audio clips (up to 3; each 2-15s, 15s combined; WAV/MP3) for sound/voice guidance, addressed as "Audio 1"… Each item is a public audio URL or an uploaded-file URL from POST /v1/uploads. Free. Can't be combined with `image`/`end_image`.
reference_imagesopt	IMAGE		Optional reference images (up to 5) for subjects/style in reference-to-video mode, addressed in the prompt as "Image 1", "Image 2"… The first two are free; each additional image adds 50 credits. Each item is a public image URL, a data URI, raw base64, or an uploaded-file URL from POST /v1/uploads. Can't be combined with `image`/`end_image`.
reference_videosopt	VIDEO		Optional reference video clips (up to 3; each 2-15s, 15s combined; MP4/MOV, ≤50MB each) for motion/identity guidance, addressed as "Video 1"… Each item is a public video URL or an uploaded-file URL from POST /v1/uploads (passed by URL, not inlined). Billed at 170 credits per second of INPUT on top of the output — estimate via POST /v1/cost. Can't be combined with `image`/`end_image`.
resolutionopt	COMBO	768p	Output resolution. Pricing varies by resolution — see `resolution_pricing` in GET /v1/models. 2k is only available when extending a `video`.
seedopt	INT	-1-1–2147483647	Seed for reproducible generation. Omit for random.
videoopt	VIDEO		Optional video to EXTEND (1.6-60s, ≤50MB, aspect ratio between 2:5 and 5:2). The model continues it for `length_seconds` of new footage guided by `prompt`, and the output is the source followed by the continuation. Provide a public video URL or an uploaded-file URL from POST /v1/uploads. Can't be combined with any other media input. The source is billed at 170 credits per second (first 15 seconds) — estimate via POST /v1/cost.
force_rerunopt	BOOLEAN	false	—

Outputs (1)

Name	Type	Description
video	VIDEO	—