Two modes, one workspace

What is AI watermark removal?

Image mode removes logos and marks via prompt and optional bounding boxes (Fal). Video mode applies a fixed rectangle blur with FFmpeg delogo—ideal for bugs and corner logos.

Why use this workspace?

Upload, preview, queue async jobs, and review history in one place—with clear split between image AI and video delogo.

Watermarked source + prompt or box → cleaned output (illustrative)

Sample source (mark overlay)
Sample mark
After removal (demo)

Video delogo: sample with watermark vs after removal (configure both URLs in admin)

Sample with watermark / bug
After FFmpeg delogo

Demo only—use the workspace above with your files. Switch tabs for FFmpeg video delogo.

Prompt-driven removal

Describe what to erase; the model inpaints using both language and pixels—good for semi-transparent text.

Optional tight boxes

Drag a rectangle on the preview; coordinates feed Fal object-removal / bbox.

Video rectangle delogo

Matches ffmpeg delogo=x:y:w:h; drag on the first-frame preview to tune the box.

Quality tiers & pricing hints

Pick a quality tier for image jobs; see rough Fal fee hints before you run (finals per workflow config).

Prompt Starters

semi-transparent platform watermark and date stamp in the bottom-right corner
circular logo in the center of the frame, keep the person’s face unchanged
horizontal banner text along the bottom edge

Frequently Asked Questions

Will it be perfectly invisible?

Complex textures or heavy compression can leave subtle traces—retry with a clearer prompt or a higher-res source.

Can it remove people or clothing?

The same model can remove objects, but altering people raises consent and policy issues—only edit media you own or are licensed to change.