Hidream O1 Image Dev Edit API Documentation

Hidream O1 Image Dev Edit API Documentation

Playground

HiDream-O1-Image Dev (Edit) is the distilled, fast-inference variant of HiDream-O1-Image in edit mode — accepts a source image plus an instruction (and optional extra references for subject-driven personalization) at up to 2K resolution, at half the price of the full model. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

HiDream-O1-Image Dev is the distilled variant of HiDream-O1-Image. The same single native model handles text-to-image, prompt-driven editing, and subject-driven personalization at up to 2K (≈2048×2048) — but tuned for fast, low-step inference so you pay less and wait less per edit. Provide one or more reference images plus an instruction; the model edits while keeping identity, composition, and lighting stable.

Why Choose This?

• Distilled for speed — Fewer denoising steps than the full HiDream-O1-Image, lower latency per edit.
• Lower cost — Half the price of the full model, ideal for batch editing, e-commerce variations, and rapid iteration.
• Identity-preserving edits — Keeps the subject, pose, and overall composition stable while applying the requested change.
• Multi-reference support — Accepts one image for plain editing, or multiple reference images for subject-driven personalization (consistent characters, products, IPs).
• High-resolution output — Up to ~2048×2048 with sharp detail and natural lighting.
• Strong instruction following — Faithfully interprets edit instructions (wardrobe, scene tweaks, object swaps, style changes, on-image text).
• Multiple output formats — JPEG, PNG, or WEBP.
• Reproducibility — Use a fixed seed to recreate exact results.

Parameters

How to Use

1. Upload your reference image(s) — drag and drop or paste publicly accessible URLs. One image for editing, multiple for subject-driven personalization.
2. Write your edit instruction — describe both the change and what to keep (e.g., “Change the jacket to black leather. Keep the person, pose, and background unchanged.”).
3. Choose size (optional) — keep the source aspect ratio or pick a new one.
4. Set output_format — JPEG for compact files, PNG for lossless, WEBP for balance.
5. Set seed (optional) — pin a number to reproduce the same edit.
6. Run — submit the request and download the edited image.

Pricing

Flat rate per edit — half the price of the full HiDream-O1-Image (Edit).

Best Use Cases

• Batch e-commerce variations — Multiple colorways or styling options from a hero shot, at scale.
• Subject-driven personalization — Pass several reference images to keep a character / product / IP consistent across new scenes.
• Wardrobe & accessory edits — Swap clothing color/style, add/remove glasses, hats, jewelry.
• Background & scene tweaks — Replace backgrounds or swap props while keeping the subject pixel-stable.
• Object add / remove / replace — Cleanly modify specific elements without disturbing the rest of the scene.
• Style transfer — Apply a new look while preserving the subject’s identity.
• Marketing iteration — Quickly produce many creative variants per source image at low cost.

Dev vs. Full — Which to Pick?

Prompting Guide

For clean results, specify both the change and what to preserve:

Template: > Keep the same person, pose, and background. Change [target edit]. Keep lighting natural and consistent.

Examples:

• “Keep the same person and pose. Change the outfit to a light gray sweater and add gold thin-rimmed glasses. Keep the background and lighting unchanged.”
• “Keep identity and composition. Replace the jacket with a black leather jacket; keep realistic fabric texture and natural shadows.”
• “Keep the subject unchanged. Remove the object on the table and keep the scene lighting consistent.”
• “Keep the person and pose. Change the background to a sunlit beach at golden hour.”

For subject-driven personalization, pass several reference images of the same subject (different angles / outfits / lighting) and describe the new scene; the model uses them jointly to keep identity consistent.

Pro Tips

• Always state what should stay the same — it dramatically improves identity preservation.
• For multi-reference personalization, give 3–5 varied shots of the subject for best identity locking.
• Reference elements explicitly: “the text on the shirt”, “the background”, “the person on the left”.
• Pin a seed while iterating so you can compare prompt variants against a fixed baseline.
• For multi-step edits, prefer sequential single-purpose prompts over one long compound instruction.
• Use png for lossless edits when output will be re-edited downstream.

Notes

• If using URLs for the source images, ensure they are publicly accessible.
• The enable_base64_output and enable_sync_mode options are only available through the API.
• Higher resolutions may slightly increase processing time.
• Ensure prompts comply with content guidelines.

Related Models

• HiDream-O1-Image Dev · Text-to-Image — Same distilled model, in text-to-image mode.
• HiDream-O1-Image · Edit — The full (non-distilled) edit model for highest fidelity.
• HiDream-O1-Image · Text-to-Image — Full model in text-to-image mode.

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"
  ],
  "size": "2048*2048",
  "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/hidream-o1-image-dev/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