Introduction
MiniMax H3 Lite reference-to-video API creates 4–15 second clips with sound at 480p or 768p, guided by image, video and audio references. Name each image, clip and voice in the prompt, and check that the cast holds together before you plan a series. Its lightweight design suits trying ideas and producing multiple variations. Every request gives you one MP4 in which picture and sound come out together.
Choose a mode by starting material
| What you have | Mode |
|---|---|
| only an idea | MiniMax H3 Lite Text to video |
| a frame to animate, with an optional closing frame | MiniMax H3 Lite Image to video |
| people, outfits, places, motion clips or voices | MiniMax H3 Lite Reference to video |
Key features & highlights
Up to 15 references
MiniMax H3 Lite accepts up to 9 images, 3 video clips (15 seconds in total) and 3 audio clips in one request. Each file can supply something different: a person, an outfit, a room, a motion or a voice.
Name them in the prompt
Point to them as "the woman in Image 1", "the movement of Video 1" and "the voice from Audio 1". Numbers follow the order of each list.
Billing that follows the inputs
Five images are covered per request and each extra image adds a small charge; reference video seconds count like output seconds, and audio is free.
Request parameters
| Parameter | Required | Description |
|---|---|---|
prompt | Yes | Required, up to 7,000 characters; anything longer is rejected. Put spoken lines in double quotes. at least 1 characters · at most 7000 characters |
reference_image_urls | No | Send up to 9 images and number them in the order the prompt uses (Image 1, Image 2...). The first 5 come free; every image after the fifth adds a small charge. at least 1 items · at most 9 items |
reference_video_urls | No | Up to 3 clips of 2 to 15 seconds each, 15 seconds in total. Their seconds are billed like output seconds. at least 1 items · at most 3 items |
reference_audio_urls | No | Up to 3 WAV or MP3 clips that set a voice or a sound, at no charge. Add at least one image or video beside them. at least 1 items · at most 3 items |
aspect_ratio | No | One of 16:9 (the default), 9:16, 1:1, 4:3, 3:4 or 21:9. Does not change the price. |
resolution | No |
|
duration_seconds | No | Whole seconds, 4 to 15, default 5. Billed per second together with the resolution. |
Pricing
Each second of output is charged at the rate of the resolution you choose, 480p or 768p. Seconds of reference video count like output seconds, the first 5 reference images are included, every further image adds a small charge, and audio costs nothing. The current rate is always shown on the model page.
| Specification | Price |
|---|---|
| $0.025 / second |
| $0.04 / second |
Use cases
Scenes assembled from separate files
A portrait, an outfit photo and a picture of a room can be sent as three images, and MiniMax H3 Lite puts that person in those clothes in that place, with a voice clip as Audio 1 if the character speaks.
The same character across clips
Send the same portrait as Image 1 in every request and MiniMax H3 Lite keeps the face and outfit while the scene changes around it, which is how a recurring character or a virtual presenter stays recognisable across a series of clips.
Copying a motion
Send a short clip as Video 1 and ask MiniMax H3 Lite to recreate its motion with the person from Image 1, for dance, gesture or camera moves. A short, clean reference clip makes the motion easy to read, and it keeps the bill down because its seconds are charged.
Prompt guide & pro tips
One job per reference
Give each file a clear role: a face, an outfit, a room, a motion or a voice, and say that role in the prompt. Mixing several roles in one picture makes the result less predictable, so a cropped face works better than a group photo.
Audio needs a picture
Audio clips cannot be the only reference; send at least one image or video, and keep the spoken words in the prompt in double quotes. The audio sets how the voice sounds, while the prompt decides what is said and who says it.
Trim reference video first
Clips are fitted to 2 to 15 seconds each and 15 seconds in total, and those seconds are billed like output, so cut the reference down to the part that shows the motion you want before you send it.
Each second of output is charged at the rate of the resolution you choose, 480p or 768p. Seconds of reference video count like output seconds, the first 5 reference images are included, every further image adds a small charge, and audio costs nothing. The current rate is always shown on the model page.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${SPICY_IDEMPOTENCY_KEY:-$(uuidgen)}"
curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
-H "Authorization: Bearer $SPICY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "minimax/h3-lite/reference-to-video",
"input": {
"aspect_ratio": "16:9",
"duration_seconds": 5,
"prompt": "A slow aerial shot drifting over a misty pine forest at sunrise",
"resolution": "768p",
"reference_image_urls": [
"https://example.com/reference.png"
]
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.
