Minimax H3 Reference-to-Video
- Supports up to 9 reference images, 3 reference videos, and 3 reference audio clips
- At least one reference image or video is required; audio-only input is not supported
- Output duration:
5–15seconds; currently only2kis supported - Asynchronous processing; use the returned task ID to query status
- Generated video URLs are valid for 24 hours; save them promptly
Authorizations
##All endpoints require Bearer Token authentication##
Get API Key:
Visit the API Key Management Page to obtain your API Key
Add to request header:
Body
Video generation model name
minimax-h3-reference-to-video "minimax-h3-reference-to-video"
Describe how the reference assets should be used and the video you want to generate.
Prompt requirements:
- Required and cannot be empty
- Chinese and English are supported
- We recommend no more than
500Chinese characters or1000English words; the model may ignore some details in an overly long prompt
Asset reference rules:
- Refer to assets as
Image 1,Image 2,Video 1,Audio 1, and so on; do not use Seedance-style@image1or@video1 - Numbering starts at
1and follows the order within each URL array - The first item in
image_urlsisImage 1, the first invideo_urlsisVideo 1, and the first inaudio_urlsisAudio 1
Input restrictions:
image_startandimage_endare not supported- Provide at least
1reference image or1reference video; reference audio cannot be used by itself
1"Replace the person in Video 1 with the character from Image 1, preserving the face, hairstyle, and clothing. Keep the motion, pacing, composition, and camera language of Video 1, and perform the dialogue with the voice from Audio 1 with synchronized lips."
Array of HTTP(S) reference-image URLs. Defaults to []; maximum 9 images.
Role and numbering:
- Every image is used as
reference_image - The first item is
Image 1, the second isImage 2, and so on
Image requirements:
- Formats: JPG, JPEG, PNG, WEBP, HEIC, HEIF
- File size: no more than
30MBeach - Width and height: each from
256to5760px - Aspect ratio (width/height):
0.4–2.5 - URLs must use HTTP(S) and be directly accessible by the service
- The complete JSON request body must not exceed
64MB; Base64 andmm_file://inputs are not accepted
Combination rules:
- Provide at least
1reference image or1reference video - Supplying only
audio_urlsreturns a parameter error
9Array of HTTP(S) reference-video URLs. Defaults to []; maximum 3 videos.
Role and numbering:
- Every video is used as
reference_video - The first item is
Video 1, the second isVideo 2, and so on
Video requirements:
- Containers: MP4 (
.mp4), MOV (.mov) - Video codecs: H.264/AVC, H.265/HEVC
- Embedded audio codecs: AAC, MP3
- File size: no more than
50MBeach - Duration:
2–15seconds per clip; total reference-video duration no more than15seconds - Width and height: each from
256to5760px - Aspect ratio (width/height):
0.4–2.5 - Frame rate:
23.976–60FPS - URLs must use HTTP(S) and be directly accessible by the service
- The complete JSON request body must not exceed
64MB; Base64 andmm_file://inputs are not accepted
Billing:
- Reference-video input duration is billable
Combination rules:
- Provide at least
1reference image or1reference video - Supplying only
audio_urlsreturns a parameter error
3Array of HTTP(S) reference-audio URLs. Defaults to []; maximum 3 clips.
Role and numbering:
- Every audio clip is used as
reference_audio - The first item is
Audio 1, the second isAudio 2, and so on
Audio requirements:
- Formats: WAV, MP3
- File size: no more than
15MBeach - Duration:
2–15seconds per clip; total reference-audio duration no more than15seconds - URLs must use HTTP(S) and be directly accessible by the service
- The complete JSON request body must not exceed
64MB; Base64 andmm_file://inputs are not accepted
Combination rules:
- Reference audio cannot be used by itself
- When using
audio_urls, also provide at least1reference image or1reference video
3Output video duration in seconds. Defaults to 5 seconds.
Value restrictions:
- Only integers from
5through15, inclusive, are supported - Decimals, numeric strings,
auto, and-1are not supported - Output duration directly affects billing
5 <= x <= 155
Output video resolution. Defaults to 2k.
Available value:
2k: the only resolution currently supported
Notes:
768pis not yet available and returns a parameter error- Output bitrate, frame rate, video codec, and audio codec are selected by the platform and are not configurable
2k "2k"
Output video aspect ratio. Defaults to adaptive.
Available values:
adaptive: the model selects a suitable ratio from the reference assets and prompt21:9: ultrawide16:9: landscape4:3: standard landscape1:1: square3:4: standard portrait9:16: portrait
adaptive, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 "adaptive"
HTTPS callback URL for task completion
Callback timing:
- Triggered when the task is completed or failed
- Sent after billing confirmation is complete
Security restrictions:
- Only HTTPS protocol is supported
- Callbacks to private IP addresses are prohibited (127.0.0.1, 10.x.x.x, 172.16-31.x.x, 192.168.x.x, etc.)
- URL length must not exceed
2048characters
Callback mechanism:
- Timeout:
10seconds - Up to
3retries after failure (at1/2/4seconds after failure respectively) - Callback response body format is consistent with the task query endpoint response format
- A 2xx status code is considered successful; other status codes trigger retries
^https://"https://your-domain.com/webhooks/video-task-completed"
Response
Video generation task created successfully
Task creation timestamp
1761313744
Task ID
"task-unified-1774857405-abc123"
Actual model name used
"minimax-h3-reference-to-video"
Specific type of the task
video.generation.task Task progress percentage (0-100)
0 <= x <= 1000
Task status
pending, processing, completed, failed "pending"
Video task details
Output type of the task
text, image, audio, video "video"
Usage and billing information