v0.8.7
Vendor-adaptive default strength
The default img2img denoising strength is now chosen from the detected SynthID vendor (C2PA issuer) instead of a single fixed value:
| Source (C2PA issuer) | Default strength |
|---|---|
| OpenAI (gpt-image / ChatGPT) | 0.10 |
| Google (Gemini / Imagen) | 0.15 |
| Unknown (metadata stripped / other) | 0.15 |
Explicit --strength always overrides.
Why: an oracle-verified June 2026 study (clean pipeline, text/face protection OFF, per-image openai.com/verify or Gemini-app verdict) found OpenAI's SynthID clears at 0.05 across 1024-1600 px (resolution-independent), while Google's is about 3x more robust and needs 0.15. The dominant factor is the vendor, not resolution. The previous single 0.30 default was overkill for OpenAI images (unnecessary quality loss). The older "large images need more strength" guidance was traced to the text-protection bug fixed in v0.8.5, not to resolution.
pip install -U remove-ai-watermarks
remove-ai-watermarks all image.png -o clean.png
Caveat: Google's 0.15 was validated on the capped --max-resolution 1536 path; a very large native Gemini image may need more (raise --strength if the oracle still detects it). See docs/synthid.md for the full data.