Introduction
MiniMax H3 LoRA Reference to video builds one shot from your reference material and up to three of your own LoRA files on the open-weights MiniMax H3. Send up to nine images, three clips and three audio clips, refer to them by number in the prompt, and get back a 3 to 15 second MP4 with a stereo soundtrack at 480p, 540p, 768p or 1080p in one of seven aspect ratios.
Choose a mode by starting material
| What you have | Mode |
|---|---|
| a written scene | MiniMax H3 LoRA Text to video |
| an opening image | MiniMax H3 LoRA Image to video |
| reference images, clips or audio | MiniMax H3 LoRA Reference to video |
Key features & highlights
Three kinds of reference in one request
Images fix a face, an outfit or a room. Clips carry motion, pacing and camera movement, and their soundtracks come along automatically. Audio clips give a voice, a beat or a room tone to follow. Address them as Picture 1, Video 1 and Audio 1 so each input has one clear job.
LoRAs and references together
A character LoRA and reference stills of the same person hold identity better than either one alone. Add a style LoRA at a lower scale and the whole shot takes on your look without losing the subject.
Uncensored across every input
Reference images are accepted as they are, whatever they show. With a LoRA built for the strongest content and our strongest prompt, the shot rendered as written with the referenced subject, and the soundtrack came back in the same file.
Request parameters
| Parameter | Required | Description |
|---|---|---|
prompt | Yes | Name each input by number and give it one job: <Picture 1> is the woman, <Picture 2> is the room, <Video 1> is the camera move, Audio 1 is her voice. at least 1 characters · at most 7000 characters |
reference_image_urls | No | Up to nine images. Each one adds a per-reference charge. at least 1 items · at most 9 items |
reference_video_urls | No | Up to three clips for motion, pacing or camera. Their seconds are billed like output seconds, capped at 15 in total. at least 1 items · at most 3 items |
reference_audio_urls | No | Up to three audio clips for a voice, a beat or ambience. Each one adds a per-reference charge. at least 1 items · at most 3 items |
aspect_ratio | No | 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 or 9:21. Does not change the price. |
resolution | No | 480p, 540p, 768p or 1080p. The per-second rate rises with each step. |
duration_seconds | No | Output length from 3 to 15 seconds. Reference clip seconds are added on top when billing. |
loras | No | Up to three {path, scale} entries trained for MiniMax H3; scale 0 to 4. A character LoRA plus reference stills of the same person gives the tightest identity hold. Default: |
seed | No | Keeps a series comparable while you swap references or scales. |
Pricing
Billed per second at the chosen resolution; reference clip seconds count as billed seconds (up to 15 in total), and each reference image or audio clip adds a per-reference charge. LoRAs add no separate charge.
| Specification | Price |
|---|---|
| $0.06 / second |
| $0.09 / second |
| $0.12 / second |
| $0.24 / second |
Use cases
A recurring character in new scenes
Keep the same two or three stills of your character, add their LoRA, and change only the setting and action lines between shots. Each new scene starts from the same identity instead of re-rolling it.
Copy a move from a clip
Send a short clip of the dance, the fight beat or the camera move you want, plus a still of the new performer, and describe the new setting. Say outright that the person comes from the picture and only the motion from the clip: in our test, leaving that sentence out brought the dancer from the clip along with the move. Keep the clip short, since its seconds are billed like output seconds.
Put a voice under the picture
Add a voice or music clip as Audio 1 and tell the model who speaks or what the scene moves to. It is the quickest way to get a line of dialogue in a voice you already have.
Prompt guide & pro tips
Open with the map
Start the prompt with one short sentence per input: Picture 1 is the woman, Picture 2 is the hotel room, Video 1 is the dolly move. Then write the action and finish with the sound. Inputs without a job tend to get blended.
Watch the reference budget
Every reference image and audio clip adds a small charge, and clip seconds are billed like output seconds up to 15 in total. The live estimate shows the full amount before you send. Nine overlapping stills cost more and usually hold identity worse than three good ones.
Let the LoRA carry the look
When a style LoRA is loaded, do not also describe that style at length in the prompt. Describe the subject and the action, and let the weights handle the rendering.
API quick start
# Generate the idempotency key once; reuse it when retrying after a lost response.
export SPICY_IDEMPOTENCY_KEY="${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: $SPICY_IDEMPOTENCY_KEY" \
-d '{
"model": "minimax/h3-lora/reference-to-video",
"input": {
"aspect_ratio": "16:9",
"duration_seconds": 5,
"prompt": "A slow aerial shot drifting over a misty pine forest at sunrise",
"resolution": "480p",
"reference_image_urls": [
"https://example.com/reference.png"
]
}
}'
# Copy data.taskId from the response into TASK_ID, then poll the status.
TASK_ID="TASK_ID_FROM_RESPONSE"
curl "https://api.spicyapi.ai/api/v1/jobs/recordInfo?taskId=$TASK_ID" \
-H "Authorization: Bearer $SPICY_API_KEY"
See the API reference for authentication, the complete schema, polling, and error handling.
