Flux Dev LoRA Ultra Fast API Documentation

Flux Dev LoRA Ultra Fast API Documentation

Playground

Ultra-fast FLUX.1 [dev] endpoint with LoRA support for high-quality image generation, personalization, and brand/style adaptation. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

FLUX Dev LoRA Ultra Fast is a high-speed image generation and editing model that supports applying up to three LoRAs in a single run. It’s built for rapid iteration: generate new images from a prompt, or guide results with an input image (image-to-image) and an optional mask for localized edits. With LoRA stacking, you can quickly dial in style, character identity, or product aesthetics while keeping latency and cost low.

Key capabilities

• Text-to-image generation with fast turnaround
• Image-to-image generation for guided edits and variations
• Masked editing (inpainting-style) for localized changes
• LoRA support: add up to 3 LoRAs per run
• Adjustable strength for how strongly the input image is preserved
• Custom output size, seed control, and standard image formats

Use cases

• Rapid style exploration using one or more LoRAs (photoreal, illustration, anime, etc.)
• Character consistency by stacking identity + style LoRAs
• Product mockups: swap materials/colors and keep composition stable
• Portrait retouching or wardrobe/background changes with a mask
• High-volume generation for ads, thumbnails, and creative testing

Pricing

Inputs

• prompt (required): what to generate or how to edit
• image (optional): source image for image-to-image
• mask_image (optional): edit region mask (for localized edits)
• loras (optional): list of LoRA items (up to 3)

Parameters

• strength: how much to follow the input image (higher = more change; lower = more preservation)
• loras[].path: LoRA identifier (owner/model) or a public.safetensors URL
• loras[].scale: LoRA weight (typical range ~0.6–1.2 depending on LoRA)
• width / height: output size
• num_inference_steps: sampling steps
• guidance_scale: prompt adherence strength
• num_images: number of images to generate per run
• seed: random seed (-1 for random; set for reproducible results)
• output_format: jpeg / png / webp, etc.
• enable_base64_output: return BASE64 instead of URL (API only)

Prompting tips

• If you’re stacking LoRAs, keep the base prompt clean and let LoRAs do the heavy lifting.
• For edits, describe the change explicitly and keep “what stays the same” implicit via the input image.
• For text rendering in images, keep on-image text short and specify placement/material.

Example prompts

• Super realism, ultra high-resolution photograph, cinematic lighting, shallow depth of field, shot on a Sony A7III, natural skin texture, sharp eyes, soft bokeh background.
• Edit: change the outfit to a light gray sweater and gold thin-rimmed glasses, keep the same face and hairstyle, realistic studio lighting.

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",
  "strength": 0.8,
  "size": "1024*1024",
  "num_inference_steps": 28,
  "guidance_scale": 3.5,
  "num_images": 1,
  "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/flux-dev-lora-ultra-fast" \
  -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