Minimax H3 Image Edit LoRA API Documentation

Minimax H3 Image Edit LoRA API Documentation

Playground

MiniMax H3 Image, the open-weights release of MiniMax’s image model, edits images from natural-language instructions with up to 9 input images and custom LoRA weights, native output from 1K to 4K. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

MiniMax H3 Image Open-Weights Edit LoRA runs the open-weights release of MiniMax H3 Image (weights published by MiniMax). It edits an existing image from a text instruction: the first input image is the one being edited, and up to 8 more can be used as references, with up to 3 custom LoRA weights per request. Output is generated natively from 1k up to 4k.

Why Choose This?

• Open-weights model — The open weights MiniMax released for H3 Image, the same model family as MiniMax H3 video.
• LoRA support — Load up to 3 LoRA weights in one request, each with its own scale.
• Instruction-based editing — Describe what should change; the first image is the editing target.
• Multi-image references — Use up to 9 images, addressed as <Picture 1>, <Picture 2>, and so on.
• Native output up to 4K — Choose 1k, 1.5k, 2k, 3k, or 4k; every tier is generated at its own size.
• Built-in prompt enhancement — Short prompts in any language are expanded into a detailed description before generation.
• Flexible framing — Square, portrait, landscape, and ultrawide aspect ratios.
• Multiple output formats — Export as JPEG, PNG, or WebP.

Parameters

Pricing

Pricing is based on the selected resolution, with a fixed $0.015 LoRA surcharge per generation included in the table.

The first input image is included in the base price. Each additional input image costs $0.01.

Additional Images

Aspect ratio, output format, and seed do not add separate charges.

Notes

• prompt and images are required.
• images supports between 1 and 9 inputs; the first is the image being edited.
• resolution defaults to 1k; output_format defaults to jpeg.
• Use seed=-1 for random seed selection.
• Up to 3 LoRA weights per request, each as {path, scale}.

Related Models

• MiniMax H3 Image Open-Weights Text-to-Image — Generate images directly from text prompts.
• MiniMax H3 Image Open-Weights Edit — Edit existing images with up to 9 input images.
• MiniMax H3 Image Open-Weights Text-to-Image LoRA — Generate images with custom LoRA weights.
• MiniMax H3 Image — The same model through MiniMax’s official API.
• P-H3-Image — Pruna AI’s image model built on MiniMax H3 open weights.

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-lora" \
  -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