T/08 · REMOVE WATERMARK AI

The mark vanishes. Everything else, untouched.

AI inpainting reconstructs pixels beneath logos, text, dates and stamps from local context. The mark goes; the rest of the image stays pixel for pixel identical to the source.

Drop a photo to remove its watermark

or drag and drop here

JPG · PNG · WebP · AVIF · up to 50 MB

S/02 — HOW IT WORKS

Three steps. No detours.

Upload, process, download. Files are auto-deleted 24 hours after processing.

S/02 · 01
compress · selected

Choose a tool

Pick what you need from the toolkit.

S/02 · 02

Drop your image

Drag, click, or paste. No upload bar required.

S/02 · 03

Download

Result is ready, original untouched.

S/SPECIMEN — INPAINT, NO TRACES

Mark out. 99.3% of the image untouched.

Source-to-output comparison. The annotated region is synthesized from local context — model trained on billions of real image patches. Every pixel outside the region is byte-equivalent to the source.

⊕ S/SPEC · STAMPED.JPG · ID 0F1C

© SAMPLE · 2026

.JPG · MARKED

source · 1 watermark

INPAINT

imgpix-erase-v1

context window · 60 px

.JPG · CLEAN

output · region inpainted

720 × 80 px region · 57 600 px synthesized · 99.3% of image untouched · imgpix-erase-v1

S/V — WHY REMOVE WATERMARKS

Inpainting, four angles.

T/08.1 · WATERMARK TYPES6 / 6 HANDLED
©STK

CORNER LOGO

© SAMPLE

DIAGONAL STAMP

2026·05·07

DATE STAMP

REPEATING

PREVIEW

TEXT OVERLAY

© RIGHTS RESERVED

COPYRIGHT BAR


Auto-detección · pincel manual opt-in cuando hace falta

TYPES

Logos, text, dates, stamps — all erased.

Doesn't matter what mark: corner badge, diagonal stripe, burn-in date, repeating pattern, copyright bar or big text overlay. The model identifies the region and reconstructs it cleanly.

T/08.2 · CONTEXT INPAINT · 16 × 10VOID 6 × 3 → FILLED

MASK · CONTEXT


VOID: 6 × 3 cells

CONTEXT: 142 cells

→ analyze surroundings

SYNTH≈ 12 s

RECONSTRUCTED


SEAMS: none

PATTERN: continued

DETAIL: synthesized


El modelo lee 60 px alrededor de la región · sintetiza relleno coherente

INPAINT

Reconstruction from local context.

The model reads pixels surrounding the mark, synthesizes a fill coherent with local texture, gradient and pattern, and stitches it without visible seams. Not erasure — learned reconstruction.

T/08.3 · DIFF MAP · MODIFIED PIXELS2.34% TOUCHED

RESULT


full color

DIFF

MODIFIED REGION


x: 600..1320 · y: 600..680

cobalt = touched · black = identical to source


TOTAL PIXELS

2 457 600

1920 × 1280

MODIFIED

57 600

2.34% · region anotada

UNCHANGED

2 400 000

97.66% · byte-equivalent

PRESERVES

Only the annotated region is touched.

Outside the mask, every pixel of the output file is byte-equivalent to the original. Same color, same profile, same resolution, same EXIF if you choose to keep it.

T/08.4 · USE CASESRESPONSIBLE USE
FAMILY

Date stamps

"2026.05.07" grabados de cámara antigua sobre fotos personales

STOCK

Licensed marks

Stock que has comprado, con la marca del proveedor por encima

SCAN

Admin watermarks

Documentos escaneados con sellos administrativos o de archivo

ARCHIVE

Old marks

Tus marcas viejas en archivos personales que quieres limpiar

RESPONSIBLE USE

Solo en imágenes que posees o tienes permiso explícito para modificar. Quitar marcas de copyright ajeno puede violar la ley.

USES

Restore what's yours.

Family photos with old-camera date stamps, licensed stock with provider marks, admin-stamped scans, personal archives with old marks of yours — all cleaned without opening Photoshop.

SPECT/08
INPUTS
JPG · PNG · WebP · AVIF · HEIC · TIFF
OUTPUT
Same format · resolution preserved
MODEL
imgpix-erase-v1 · context inpainting
MASK
Auto-detect · manual brush opt-in
LATENCY
6–18 s typical, depending on size and region
PRIVACY
No training on your images

S/F — FAQ

Remove watermark, briefly answered.

How does the AI fill the missing area?
It analyzes color, texture and pattern from pixels surrounding the mark and synthesizes a plausible reconstruction. For repetitive backgrounds (sky, grass, fabric) results are indistinguishable from the original. Over complex textures it may leave subtle traces on close inspection.
Are results always perfect?
No. Simple marks on uniform backgrounds come out perfect. Complex marks over detailed textures may leave traces — visible only under zoom. If the result isn't convincing, refine the mask manually and re-run.
Is it legal to remove watermarks?
Only on images you own or have explicit permission to modify. Removing marks from third-party copyrighted material can violate copyright law and platform terms. Use it responsibly.
Are my images used to train the model?
No. Your uploads are used exclusively to process your job. Zero feedback into the training pipeline. The model is frozen and versioned.