Qwen Image Edit API Documentation

Qwen Image Edit API Documentation

Playground

Qwen-Image-Edit is a 20B MMDiT image-to-image model offering precise bilingual (Chinese & English) text edits while preserving style. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

A next-gen image editing model built on Qwen-Image 20B MMDiT. Qwen-Image-Edit delivers precise bilingual text editing (Chinese & English), supports both semantic and appearance-level edits, and preserves the original style — making it one of the most capable image editing models available.

Why It Looks Great

• Dual-mode editing: Supports both appearance-level edits (add/remove/modify elements while keeping other regions pixel-accurate) and semantic-level edits (IP creation, rotation, style transfer with global coherence).
• Precise text editing (CN/EN): Edit on-image text directly — add, delete, or replace while retaining the original font, size, kerning, and style.
• Style preservation: Maintains palette, lighting, brushwork, and overall look even under substantial edits.
• 20B parameter power: Built on the robust Qwen-Image foundation for superior understanding and generation.
• State-of-the-art performance: Achieves SOTA results across multiple public image editing benchmarks.
• Prompt Enhancer: Built-in tool to refine your editing instructions automatically.

Parameters

How to Use

1. Write your edit instruction — describe what you want to change in the image.
2. Use Prompt Enhancer (optional) — click to refine your editing instruction.
3. Upload your image — drag and drop or paste a public URL.
4. Set dimensions (optional) — adjust width and height for the output size.
5. Set seed (optional) — use -1 for random, or a specific number to reproduce results.
6. Choose output format — select jpeg or png.
7. Run — click the button to apply the edit.
8. Download — preview and save your edited image.

Pricing

Flat rate per edit.

Best Use Cases

• Text Replacement — Change text on signs, labels, clothing, or any surface in images (supports Chinese & English).
• Element Modification — Add, remove, or swap objects while preserving untouched areas pixel-perfectly.
• Style Transfer — Apply new artistic styles while maintaining semantic content.
• IP Creation — Generate character variations and creative adaptations.
• Object Manipulation — Rotate, reposition, or transform objects within scenes.
• Product Customization — Create variations of product images with different details or branding.

Example Prompts

• “Change the background to daytime”
• “Replace the text on the sign to say ‘Welcome’”
• “Remove the person on the left side of the image”
• “Change the dress color from red to blue”
• “Add sunglasses to the person”

Editing Modes Explained

Pro Tips for Best Results

• Be specific about what to change and what the result should be.
• For text editing, the model preserves original typography — just specify the new text content.
• Reference elements clearly: “the text on the shirt”, “the background”, “the person’s hair”.
• For appearance edits, unchanged areas remain pixel-perfect — focus only on what needs editing.
• Use Chinese prompts for editing Chinese text in images for best accuracy.
• The model understands context — describe the intent, not just the mechanical change.

Notes

• If using a URL for the image, ensure it is publicly accessible. A preview thumbnail confirms successful loading.
• The enable_base64_output and enable_sync_mode options are only available through the API.
• Complex edits may require more specific prompts for accurate results.
• For multiple edits to the same image, consider processing sequentially for better control.

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",
  "image": "https://interactive-examples.mdn.mozilla.net/media/cc0-images/painted-hand-298-332.jpg",
  "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/qwen-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