Black Forest Labs

FLUX Tools

Choose Erase for masked image cleanup, Outpainting for extending a composition, Virtual Try-On for combining person and garment images, or Deblur for recovering a soft source. Video Upscale enlarges an existing clip by 1.5x to 3x using Precise or Creative mode, preserving aspect ratio and audio. Select the tool before preparing its source media.

How to use this model

  1. 1

    Choose the tool for your task: Erase, Outpainting, Virtual Try-On, Deblur, or Video Upscale.

  2. 2

    Upload the highest-quality source available; avoid resizing it before masking or expansion.

  3. 3

    For Erase, paint the complete unwanted area in white on a same-size mask and keep everything else black.

  4. 4

    For Outpainting, set the new canvas and offsets; for Virtual Try-On, provide clear person and garment images.

  5. 5

    Inspect seams, repeated textures, garment boundaries, and facial detail at full size before saving the result.

  6. 6

    For Video Upscale, upload a clip up to 20 seconds and 2K source resolution, choose the scale and mode, and optionally describe its content.

ToolsAvailable

FLUX Erase

FLUX Erase removes an object or blemish from a source image and fills the selected region from surrounding context. The mask must match the source dimensions: white marks what to remove and black protects the rest. It is intended for localized cleanup, not global restyling.

Try now

Properties

  • Requires a source image and a same-size removal mask.
  • White mask pixels are removed; black pixels are preserved.

Best for

  • Object removal
  • Image cleanup

Tips

  • The mask must match the source dimensions.
  • Paint the area to remove white and preserve the rest in black.

How to use this model

  1. 1

    Upload the source at its original dimensions.

  2. 2

    Create a same-size black mask and paint the complete unwanted area white.

  3. 3

    Include a small margin around difficult edges without covering content that must remain.

  4. 4

    Generate and inspect reconstructed texture, straight lines, and object boundaries.

  5. 5

    Refine the mask and rerun if remnants or obvious fill patterns remain.

Examples

Example output from FLUX Erase
Beach scene before the camper was erased
100%
A reviewed image output created with FLUX Erase in Omnipix.

Parameters

Source Image

image
Required

Removal Mask

mask
Required

White pixels are removed; black pixels are preserved. Use the same dimensions as the source image.

Mask Dilation

dilate_pixels
Optional
Default: 10Range: 0 – 25

Seed

seed
Optional
ToolsAvailable

FLUX Outpainting

FLUX Outpainting places the source on a larger canvas and synthesizes the missing area around it. Horizontal and vertical offsets control where the original sits, while optional guidance can describe the extended scene. It is useful for reframing, changing aspect ratio, and adding composition space without stretching the original.

Try now

Properties

  • Expands a source image onto a larger canvas.
  • Supports source placement with horizontal and vertical offsets.

Best for

  • Canvas expansion
  • Background extension

Tips

  • Use offsets to control where the source sits on the output canvas.
  • High mode prioritizes fidelity; fast mode prioritizes latency.

How to use this model

  1. 1

    Upload the uncropped source at the best available quality.

  2. 2

    Set output width and height for the intended composition.

  3. 3

    Use horizontal and vertical offsets to position the source on the new canvas.

  4. 4

    Add concise guidance for scenery, lighting, or objects that should continue outside the original frame.

  5. 5

    Check seams, perspective, horizons, and repeated elements along every expanded edge.

Examples

Camper scene expanded beyond the original frame with FLUX Outpainting
Original camper scene before expanding the canvas
100%
Drag the divider to compare the original 1360×768 frame with the expanded 1680×944 result.

Parameters

Source Image

input_image
Required

Image to place on the expanded canvas.

Output Width

width
Required
Default: 1920Range: 64 – +∞

Output Height

height
Required
Default: 1080Range: 64 – +∞

Mode

mode
Optional
Default: high

  • high

  • fast

Auto Crop

auto_crop
Optional
Default: false

Prompt

prompt
Optional

Optional experimental guidance for the extended area.

Horizontal Offset

reference_offset_x
Optional

Vertical Offset

reference_offset_y
Optional
Video UpscaleAvailable

FLUX Video Upscale

FLUX Video Upscale enlarges source video by 1.5x to 3x while preserving its aspect ratio and audio. Precise mode keeps faces and products closer to the original; Creative mode adds detail more freely. Larger sources may reach the output-size cap before the requested scale.

Try now

Properties

  • Enhance → Video - Upscale, with video output and preserved audio.
  • Precise and Creative modes; Creative is the default.
  • Output follows the source aspect ratio and is capped at 13.75 megapixels per frame.
  • Pricing depends on output frame size, clip duration, and mode; final billing uses the delivered result.

How to use this model

  1. 1

    Upload an MP4 clip up to 20 seconds, 50 MB, and 2560x1440 source resolution.

  2. 2

    Choose an upscale factor from 1.5x to 3x. The default is 2x.

  3. 3

    Use Precise for faces and products, or Creative for generated footage, textures, and scenery.

  4. 4

    Optionally describe the content to guide enhanced detail, then review the duration and quote before running.

Examples

Precise sharpens the ice, foliage and butterfly while staying close to the winter edit. 414×234 → 1242×702.
Creative adds richer frost and foliage textures, reimagining fine detail in the winter scene. 414×234 → 1242×702.

Parameters

Source Video

video
Required

MP4 up to 20 seconds, 50 MB, and 2560x1440 source resolution.

Upscale Factor

upscale_factor
Optional

Scale each side by 1.5x to 3x. Large sources may reach the output-size cap before the requested factor.

Default: 2Range: 1.5 – 3

Mode

creativity
Optional

Precise stays faithful to faces and products. Creative adds detail more freely and may change identity.

Default: 1

  • Precise

  • Creative

Content Description

prompt
Optional

Optional description to guide enhanced detail, up to 4,096 characters in Omnipix. Leave empty for a neutral upscale.

Default:
ToolsAvailable

FLUX Virtual Try-On

FLUX Virtual Try-On combines a person image with a separate garment image to render the clothing on that person. Clear body visibility and a complete, unobstructed product image give the model the strongest information about fit and construction. The result is a visualization and should not be treated as a precise sizing guarantee.

Try now

Properties

  • Requires separate person and garment images.
  • Designed for garment visualization rather than exact fit measurement.

Best for

  • Fashion visualization
  • Garment previews

Tips

  • Use a clear full-body or half-body person image.
  • Use a clean garment image with the full item visible.

How to use this model

  1. 1

    Upload a clear full-body or half-body person image with minimal occlusion.

  2. 2

    Upload one clean garment image with the entire item visible.

  3. 3

    Prefer a garment category and viewing angle compatible with the person's pose.

  4. 4

    Generate and inspect neckline, sleeves, hands, waist, hems, and garment graphics.

  5. 5

    Try cleaner inputs or a less obstructed pose if anatomy or garment boundaries distort.

Examples

Synthetic character wearing the referenced green hoodie after FLUX Virtual Try-On
Synthetic GPT Image 2 character before virtual try-on
100%

Reference

Green hoodie supplied as the garment reference
The person source was generated with GPT Image 2. FLUX Virtual Try-On applied the separate hoodie reference shown beside the comparison.

Parameters

Person

person
Required

Garment

garment
Required

Prompt

prompt
Required
Default: TRY-ON: The person of image 1 wearing the garments of image 2.

Seed

seed
Optional
ToolsRetired

FLUX Deblur

FLUX Deblur reconstructs detail in images softened by motion, defocus, or general blur. Because reconstruction can infer information that was not legible in the source, important faces, text, and product details should be checked carefully. A seed can be reused when a reproducible result is needed.

Properties

  • Targets motion blur, defocus, and general softness.
  • May reconstruct details that were not fully visible in the source.

Best for

  • Motion blur
  • Soft or defocused images

Tips

  • Use a seed when you need reproducible output.

Capabilities

  • Runtime category for restoration: Image - DeblurRuntime contract

How to use this model

  1. 1

    Upload the least-compressed version of the blurred image.

  2. 2

    Inspect whether blur affects the whole frame or a moving subject before processing.

  3. 3

    Run the deblur without prior aggressive sharpening or denoising.

  4. 4

    Compare faces, text, hair, fine edges, and repeating texture with the source at 100%.

  5. 5

    Reuse the seed for a stable rerun or try another result if reconstructed detail is implausible.

Parameters

Source Image

image
Required

Seed

seed
Optional