# aloshdenny/reverse-SynthID

reverse engineering Gemini's SynthID detection

Repository: https://github.com/aloshdenny/reverse-SynthID
Canonical: https://ross.abutalabs.com/products/reverse-synthid
Homepage: https://vt-0xff.github.io/SynthID-Explained/
Language: Python
License: NOASSERTION
License Family: other
Last push: 2026-07-17T17:33:48+00:00

## Health v2 (maintenance only)
Score: 66/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 93, release rhythm 60, longevity 18
- inputs: {"age_days": 260, "days_push": 47, "days_rel": 132, "gap_med": 2, "n_releases_24m": 2}
- flags: prerelease_only, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4815, forks 506 (observed 2026-08-28T04:09:00.322339+00:00)

## What it is
A research tool that reverse-engineers Google's SynthID watermark embedded in Gemini-generated images using spectral analysis and signal processing. It provides a detector for SynthID watermarks and multi-stage bypass attacks, with a drag-and-drop desktop GUI for the V3 bypass.

## Use cases
- detect whether an image contains a Google SynthID watermark
- remove SynthID watermarks from Gemini-generated images
- analyze the spectral signature of AI image watermarks
- study how SynthID's carrier frequency structure works
- bypass AI image watermark detection while preserving image quality
- calibrate watermark removal strength against Gemini's detector

## When to choose
- you need to detect or remove SynthID watermarks from Gemini images
- you are researching AI watermarking robustness and steganalysis
- you want a GUI-based watermark bypass without command-line work

## When to avoid
- you need to watermark your own generated images rather than remove watermarks
- you need a production-grade, legally cleared content-provenance solution
- you target watermarks from non-Google generators like DALL-E or Stable Diffusion

## Facets
- artifact type: application
- maturity: active
- function: image-processing, security, reverse-engineering, computer-vision
- domain: security, reverse-engineering, image-processing, artificial-intelligence
- platform: python, cross-platform
- tags: synthid, watermark-removal, watermark-detection, spectral-analysis, ai-watermarking, gemini, signal-processing, research, desktop

## Member repositories
- aloshdenny/reverse-SynthID (main) score 66

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:00.322339+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:18:34.514138+00:00, confidence not recorded.
  - readme: https://github.com/aloshdenny/reverse-SynthID (fetched 2026-08-28T04:09:00.322339+00:00, sha 2682112f0595)
  - homepage: https://vt-0xff.github.io/SynthID-Explained/ (fetched 2026-08-29T09:01:30.813326+00:00, sha c61f47aed208)
- Data as of 2026-08-30T08:39:29.467469+00:00.