Flux Kontext Max

Flux Kontext Max

Playground

FLUX.1 Kontext [max] boosts prompt adherence and typography generation for consistent, high-quality image editing at speed. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

FLUX Kontext Max (Image-to-Image Editing) — wavespeed-ai/flux-kontext-max

FLUX Kontext Max is a premium image-to-image editing model built for high-fidelity, instruction-following transformations. Provide a source image plus a natural-language edit prompt, and it performs precise local or global edits while maintaining strong visual coherence—ideal for demanding creative direction, high-end retouching, and style-driven transformations.

Key capabilities

• High-fidelity instruction-based image editing from a single input image
• Strong prompt adherence for complex, multi-constraint edits
• Handles both local edits (specific elements) and global edits (overall look)
• Excellent for style transformations (e.g., toy style, clay, illustration) while preserving composition

Typical use cases

• Premium retouching: lighting correction, cleanup, detail enhancement
• Background swaps with consistent lighting/shadows
• Product and branding edits requiring high accuracy
• Style transformations with minimal drift (toy, clay, cinematic, illustration)
• Creative iterations where output quality matters more than speed

Pricing

$0.08 per image.

Inputs and outputs

Input:

• image (required): Source image (upload or public URL)
• prompt (required): Edit instruction

Output:

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

Parameters

• prompt (required): Edit instruction describing what to change and what to preserve
• image (required): Source image
• seed: Fixed value for reproducibility; leave empty/random for variation
• guidance_scale: Prompt adherence strength (higher = stricter; too high may over-edit)
• aspect_ratio: Output aspect ratio (choose to control framing/cropping)

Prompting guide

For best control, use a “preserve + edit + constraints” structure:

Template: Keep [what must stay]. Change [what to edit]. Ensure [constraints]. Match [lighting/shadows/perspective].

Example prompts

• Keep the person’s face, pose, and clothing unchanged. Convert the entire image to a high-quality toy style with realistic plastic texture, soft studio lighting, and clean highlights. Keep the background composition consistent.
• Keep the subject identity and expression unchanged. Replace the background with a clean pastel studio backdrop. Match lighting direction and shadow softness.
• Remove background clutter and keep the main subject sharp. Apply a gentle cinematic color grade without changing composition.

Best practices

• Start with one change per run, then iterate for precision.
• If the edit is too strong, lower guidance_scale and add a clearer preserve clause.
• Fix seed for stable comparisons across prompt variants.
• Choose aspect_ratio intentionally to avoid unexpected cropping.

# Submit the task
curl --location --request POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/flux-kontext-max" \
--header "Content-Type: application/json" \
--header "Authorization: Bearer ${WAVESPEED_API_KEY}" \
--data-raw '{
    "guidance_scale": 3.5,
    "enable_sync_mode": false
}'

# Get the result
curl --location --request GET "https://api.wavespeed.ai/api/v3/predictions/${requestId}/result" \
--header "Authorization: Bearer ${WAVESPEED_API_KEY}"