Minimax H3 Image Edit API Documentation

Minimax H3 Image Edit API Documentation

Playground

MiniMax H3 Open Weights Image Edit re-renders a subject from up to 9 reference images into a new scene, outfit or style from a text instruction, preserving identity at 1K, 1.5K or 2K resolution. Ready-to-use REST inference API, best performance, no cold starts, affordable pricing.

Features

Run the open-weights edition of MiniMax H3 as a reference-guided image editor on WaveSpeedAI’s own GPU infrastructure. Provide 1 to 9 reference images and an instruction; the model re-renders the subject into a new scene, outfit, pose or art style while preserving identity, at 1k or 2k resolution.

Why Choose This?

• Strong identity preservation
  Faces, hairstyles, accessories and clothing details carry over from the references.

• Scene, outfit and style changes
  Move a person to a new location, change what they wear, or restyle the image as a painting or animation.

• Multi-image references
  Combine up to 9 references, addressed in the prompt as <Picture 1>, <Picture 2>, and so on.

• Aspect ratio follows your input
  By default the output keeps the aspect ratio of the first reference image; override it with aspect_ratio.

Parameters

How to Use

1. Upload references — One clear image of the subject is enough; add more for outfits, props or additional people.
2. Describe the new image — State the new scene, outfit, pose and style explicitly, e.g. The same woman from <Picture 1>, now wearing a red leather jacket, standing in a neon-lit Tokyo alley at night. Photorealistic still.
3. Choose resolution — 1k for iteration, 1.5k for a balance of detail and cost, 2k for final assets.
4. Submit — Receive the edited image URL.

Pricing

Pricing is based on the selected resolution.

The prices below include the first input image. Each additional input image costs $0.005, regardless of resolution.

Best Use Cases

• Virtual try-on and outfit swaps — Keep the person, change the clothes.
• Scene relocation — Put a subject into a new environment with matching light.
• Style transfer — Oil painting, anime or illustration versions of a photo.
• Character sheets — Consistent renders of one character across settings.

Pro Tips

• Describe the new scene fully instead of only naming the reference; short prompts that just say “the person from <Picture 1>” tend to reproduce the reference.
• For style transfer, say so explicitly, e.g. This is a painting, not a photograph.
• Keep the subject’s key features in the prompt (hair, glasses, beard) when they matter.
• Fix the seed to iterate on wording without changing the composition.

Related Models

• MiniMax H3 Text-to-Image — Generate stills from text with the same model.
• MiniMax H3 Reference-to-Video — Animate reference images into video with native audio.

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",
  "images": [
    "https://interactive-examples.mdn.mozilla.net/media/cc0-images/painted-hand-298-332.jpg"
  ],
  "aspect_ratio": "1:1",
  "resolution": "1k",
  "output_format": "jpeg"
}
JSON
)

# 1. Submit the prediction once. Never retry this request automatically:
# a retry after an unclear failure can create a second, billed task.
CHECK='The task may already exist. Check History or the predictions list before submitting again.'
if ! SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  --connect-timeout 10 --max-time 60 \
  -X POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/minimax-h3/image-edit" \
  -H "Authorization: Bearer ${WAVESPEED_API_KEY}" \
  -H "Content-Type: application/json" \
  -d "${REQUEST_BODY}"); then
  printf 'Submit failed. %s\n' "${CHECK}" >&2
  exit 1
fi

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 "${PREDICTION_ID}" ]; then
  printf 'Submission response did not contain a prediction id. %s\n' "${CHECK}" >&2
  exit 1
fi
RESULT_URL="https://api.wavespeed.ai/api/v3/predictions/${PREDICTION_ID}/result"

# 2. Poll until the prediction reaches a final status.
BODY_FILE=$(mktemp)
trap 'rm -f "${BODY_FILE}"' EXIT
while true; do
  # Transient errors (429, 5xx) are retried, honoring Retry-After.
  HTTP_CODE=$(curl --silent --show-error --fail-with-body \
    --connect-timeout 10 --max-time 60 --retry 5 --retry-delay 2 \
    --output "${BODY_FILE}" --write-out '%{http_code}' \
    "${RESULT_URL}" \
    -H "Authorization: Bearer ${WAVESPEED_API_KEY}" || true)
  case "${HTTP_CODE}" in
    2??) ;;
    000|429|5??)  # dropped connection or temporary error: keep polling
      printf 'Task %s: HTTP %s, polling again in 10s\n' "${PREDICTION_ID}" "${HTTP_CODE}" >&2
      sleep 10; continue ;;
    *) printf 'Task %s: HTTP %s\n' "${PREDICTION_ID}" "${HTTP_CODE}" >&2
      cat "${BODY_FILE}" >&2; exit 1 ;;
  esac
  RESULT=$(jq 'if type == "object" and has("data") then .data else . end' "${BODY_FILE}")
  STATUS=$(printf '%s' "${RESULT}" | jq -r '.status // empty')

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

Parameters

Task Submission Parameters

Request Parameters

Every model also accepts the global request options enable_sync_mode and enable_base64_output; see their pages for details.

Response Parameters

Result Request Parameters

Result Response Parameters