Flux Kontext Max Multi API Documentation

Flux Kontext Max Multi API Documentation

Playground

Experimental FLUX.1 Kontext [max] (multi) supports multi-image context handling for combined inputs. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

FLUX Kontext Max Multi is a high-end multi-image model for context-rich generation and editing. Provide a text prompt plus up to 3 reference images, and the model uses them as visual grounding to improve identity consistency, style matching, and scene coherence—ideal for premium creative work where one image is not enough.

Key capabilities

• Multi-image contextual generation with up to 3 reference images
• Strong identity and style consistency by grounding outputs in references
• Handles complex scenes and cinematic composition with high detail
• Great for iterative workflows: refine results while keeping the same visual target

Typical use cases

• Character consistency using multiple portraits/outfits/angles
• Product and branding consistency (packaging + logo + lighting references)
• Style steering with multiple exemplars (art style + texture + lighting mood)
• Scene creation or recomposition guided by reference frames
• High-fidelity creative direction for storyboards and marketing visuals

Pricing

$0.08 per image.

Total cost = num_images × $0.08 Example: num_images = 4 → $0.32

Inputs and outputs

Input:

• prompt (required): The generation or edit instruction
• images (required): Up to 3 reference images (upload or public URLs)

Output:

• One or more generated images (controlled by num_images, if available in your interface)

Parameters

• prompt (required): Instruction describing what to generate and how to use references
• images (required): Up to 3 reference images
• aspect_ratio: Output aspect ratio (e.g., 16:9, 1:1, 9:16)

Prompting guide (multi-reference)

Assign roles to your references to reduce ambiguity:

Template: Use image 1 for [identity]. Use image 2 for [outfit]. Use image 3 for [style]. Generate [shot description]. Keep [constraints].

Example prompts

• Use image 1 for the face identity, image 2 for outfit, image 3 for illustration style. Create a 16:9 cinematic medium shot in a rainy city street at night, neon reflections, shallow depth of field.
• Use images 1–2 to keep the same person identity from different angles. Generate a clean studio portrait with softbox lighting, neutral background, natural skin texture.
• Use image 2 for lighting mood (sunset) and image 3 for environment. Keep the subject identity from image 1 and maintain consistent color palette.

Best practices

• Use high-quality references: sharp subjects, minimal occlusion, clear lighting.
• Avoid conflicting references (e.g., drastically different styles) unless you explicitly say which one dominates.
• Let the references do most of the steering.
• Pick an aspect_ratio that matches your target layout to avoid awkward cropping.

Notes

• If an output is flagged as NSFW, it is returned as a black image with the same dimensions.

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": "21:9"
}
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-kontext-max/multi" \
  -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