Minimax H3 Singularity Reference To Video API Documentation

Minimax H3 Singularity Reference To Video API Documentation

Playground

MiniMax H3 Singularity Open Weights Reference to Video generates coherent 480P / 540P / 768P videos from prompts and multimodal references, guided by up to 9 reference images, 3 reference videos, and 3 reference audios, with native stereo audio and flexible reference-based video generation on WaveSpeedAI infrastructure. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

Run the open-weights edition of MiniMax H3 Singularity on WaveSpeedAI’s own GPU infrastructure. This endpoint is independently hosted and provides 480p, 540p, 768p, and 1080p output tiers with resolution-based per-second pricing.

Why Choose This?

• Multimodal reference-to-video generation
  Guide video generation with images, videos, audio, or combinations of multiple reference types.

• Up to 9 reference images
  Use multiple visual references for characters, objects, environments, composition, or style.

• Up to 3 reference videos
  Provide motion and video context through multiple reference clips.

• Reference video audio support
  Soundtracks from reference videos are used automatically during generation.

• Standalone audio references
  Add up to 3 separate audio references for additional voice, music, or sound guidance.

• Native audio generation
  Audio is generated together with the video rather than requiring a separate generation step.

• Four resolution tiers
  Choose 480p, 540p, 768p, or 1080p depending on cost and output requirements.

• Up to 15 seconds of generated video
  Generate clips from 3 to 15 seconds.

How to Use References

Address references directly in the prompt using angle-bracket tags:

• Images: <Picture 1> through <Picture 9>
• Videos: <Video 1> through <Video 3>
• Standalone audio: <Audio 1> through <Audio 3>

Tags should be written with the angle brackets exactly as shown.

Assigning a clear purpose to each reference helps make the generation instruction easier to interpret.

For example:

> Use the character from <Picture 1>, place them in the environment from <Picture 2>, and follow the camera movement of <Video 1>.

Reference numbering follows the order in which inputs are supplied within each reference type.

Reference video soundtracks are used automatically.

How to Write a Strong Prompt

For more controlled generation, structure the prompt around the scene progression:

1. Define references — State which image, video, or audio reference should guide each part of the generation.
2. Describe the timeline — Explain how the action develops throughout the clip.
3. Specify camera movement — Describe pans, tracking, zooms, or static camera behavior.
4. Describe audio direction — Explain the desired dialogue, music, ambience, or sound behavior.
5. Specify visible text carefully — Put exact on-screen wording in quotes when readable text matters.

Example:

> Use the character from <Picture 1> and the environment from <Picture 2>.
> [0s-3s] The character stands on the balcony while the camera slowly moves forward.
> [3s-6s] They turn toward the camera and smile.
> Match the camera movement of <Video 1>.
> Audio: quiet city ambience with natural environmental sound.

Parameters

Pricing

The generated output video and reference-video duration use the same resolution-dependent per-second rate.

Reference Pricing

Total reference-video duration is capped at 15 seconds for billing and rounded up to the next whole second.

Standalone reference_audios are charged by count, not duration.

Example

Best Use Cases

• Character consistency — Use reference images to guide recurring characters or subjects.
• Motion guidance — Use reference videos to guide movement, action, or camera behavior.
• Voice and audio reference workflows — Add standalone audio references alongside visual guidance.
• Multi-reference storytelling — Combine characters, environments, motion, and audio from different inputs.
• Cinematic video generation — Coordinate visual references, camera direction, timeline, and native audio in one generation.
• Creative prototyping — Experiment with combinations of visual, motion, and audio references.

Pro Tips

• Explicitly assign a purpose to each reference in the prompt.
• Use tags such as <Picture 1> and <Video 1> instead of vague descriptions like “the first image.”
• Keep the timeline clear when multiple actions need to happen in sequence.
• Use reference videos when motion or camera behavior matters more than a static visual reference.
• Keep total reference-video content focused because only up to 15 seconds is used.
• Use 480p for lower-cost testing before moving to 768p or 1080p.
• Set a fixed seed when comparing prompt or reference changes.

Notes

• prompt is the only schema-required parameter.
• Up to 9 reference images are supported.
• Up to 3 reference videos are supported.
• Up to 3 standalone reference audio inputs are supported.
• Total reference-video duration is limited to 15 seconds.
• Each standalone reference audio input is trimmed to a maximum of 15 seconds.
• Reference video soundtracks are used automatically.
• duration supports integer values from 3 to 15.
• resolution defaults to 480p.
• aspect_ratio defaults to 16:9.
• seed=-1 selects a random seed.

Related Models

• MiniMax H3 Singularity Reference-to-Video — Generate video with multimodal image, video, and audio reference guidance.
• MiniMax H3 Singularity Reference-to-Video LoRA — Use the Singularity reference-to-video workflow with LoRA-based customization.

Authentication

For authentication details, please refer to the Authentication Guide.

API Endpoints

Submit Task & Query Result

set -euo pipefail

export WAVESPEED_API_KEY="your-api-key"

REQUEST_BODY=$(cat <<'JSON'
{
  "prompt": "A cinematic ocean wave at sunrise, highly detailed",
  "aspect_ratio": "16:9",
  "resolution": "480p",
  "duration": 5
}
JSON
)

# 1. Submit the prediction.
SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/minimax-h3-singularity/reference-to-video" \
  -H "Authorization: Bearer ${WAVESPEED_API_KEY}" \
  -H "Content-Type: application/json" \
  -d "${REQUEST_BODY}")

TASK=$(printf '%s' "${SUBMIT_RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
PREDICTION_ID=$(printf '%s' "${TASK}" | jq -r '.id // empty')
if [ -z "${PREDICTION_ID}" ]; then
  printf 'Submission response did not contain a prediction id
' >&2
  exit 1
fi
RESULT_URL="https://api.wavespeed.ai/api/v3/predictions/${PREDICTION_ID}/result"

# 2. Poll until the prediction finishes.
while true; do
  RESPONSE=$(curl --silent --show-error --fail-with-body \
    "${RESULT_URL}" \
    -H "Authorization: Bearer ${WAVESPEED_API_KEY}")
  RESULT=$(printf '%s' "${RESPONSE}" | jq 'if type == "object" and has("data") then .data else . end')
  STATUS=$(printf '%s' "${RESULT}" | jq -r '.status // empty')

  case "${STATUS}" in
    completed) printf '%s\n' "${RESULT}" | jq '.outputs'; break ;;
    failed|cancelled|timeout|deleted) printf '%s\n' "${RESULT}" | jq . >&2; exit 1 ;;
    *) sleep 2 ;;
  esac
done

Parameters

Task Submission Parameters

Request Parameters

Response Parameters

Result Request Parameters

Result Response Parameters