Last commit added `_color_match` which shifts the face crop's mean to the
canvas mean -- the old test fed a uniform face (210) into a uniform cleaned
canvas (90), so after color-match the face was uniform 90 and the
composite was undetectable by value. Switched the fake pipeline to a
gradient face so the color-match preserves variance, and the assertion
now checks that the face region has non-zero std (composite injected
gradient pixels) instead of a value threshold.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Per the 2026-06-08 deep-research synthesis (docs/synthid-robust-identity-
research-2026-06-08.md), the entire ArcFace-class identity-adapter ecosystem
for SDXL is blocked from commercial use by InsightFace's non-commercial model
packs (antelopev2 / buffalo_l). No commercial-safe ArcFace-grade identity
stack exists today. The user explicitly opted into shipping a non-commercial
restore path (research / personal use; raiw.cc must NOT install the extra).
Architectural choice: InstantID over PhotoMaker-V2 as the default.
- PhotoMaker-V2 (CLIP+ArcFace dual encoder, txt2img only): documented upstream
identity drift on Asian male faces, visually confirmed in our cert sweep
(tatsunari rendered as a generic woman; group photo collapsed into a
patchwork).
- InstantID (ArcFace cross-attention + landmark ControlNet): semantic
identity branch + spatial weak landmark control, decoupled. Per InstantID
paper (arXiv:2401.07519) and the research report, stronger identity fidelity
on single portraits. Critically: NO original face pixels enter the diffusion
(ArcFace embedding is semantic, landmark stick figure is pure geometry), so
SynthID is not transported.
Implementation:
- New `src/remove_ai_watermarks/instantid_restore.py` mirrors the
`photomaker_restore.py` shape (lazy singletons for pipeline + FaceAnalysis,
per-face crop + _composite_faces from photomaker_restore). Loads the
InstantID community pipeline via `DiffusionPipeline.from_pretrained(
custom_pipeline="pipeline_stable_diffusion_xl_instantid")` -- no upstream
Python package needed; diffusers fetches the file from its community
examples.
- New `instantid` extra in pyproject (insightface + onnxruntime +
huggingface-hub). NON-COMMERCIAL block in the comment explains why.
- CLI: `--restore-faces-method [instantid|photomaker]`, default `instantid`.
Both methods explicitly labeled NON-COMMERCIAL in the help text.
- Engine: dispatch on `restore_faces_method` to either
`_restore_faces_instantid` or `_restore_faces_photomaker`.
- 9 control-flow tests for InstantID without model download (mirror the
photomaker_restore.py test pattern + draw_kps helper checks). 587/587 pass.
Diffusers-0.38 compat verified by upstream code inspection: the InstantID
pipeline inherits from `StableDiffusionXLControlNetPipeline`, uses only
public diffusers APIs (`encode_prompt`, `prepare_image`, `prepare_latents`,
`get_guidance_scale_embedding`), uses legacy attention processor API which
diffusers preserves for backward compat. No PhotoMaker-V1-style internal
text_encoder access. End-to-end execution will be validated by the Modal
cert sweep in the next step.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The previous commit added a real call into FaceAnalysis2 / analyze_faces inside
restore_faces_photomaker, which broke the model-free control-flow test. Stub it:
- monkeypatch _get_face_analyser to return a sentinel
- install a fake `photomaker` module with analyze_faces returning a single
512-d zero embedding
- add dtype=torch.float32 to the fake pipeline class so .to(device, dtype=...) works
11/11 green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Visual review of the GFPGAN-on-cleaned output (9-face grid, 1448x1086) showed it
only polished the already-drifted face without restoring identity — useless for the
"restore who is in the photo" intent. Dropping it.
The shipped restore path is now PhotoMaker-V2, which delivers true identity-from-
embedding face regeneration via a CLIP+ArcFace dual encoder. The ArcFace branch
pulls InsightFace antelopev2/buffalo_l model packs at runtime, which InsightFace
releases under a research-only license, so the whole extra is **NON-COMMERCIAL**.
raiw.cc and any monetized deployment must NOT install the `photomaker` extra.
This is called out at every entry point: CLI flag help, module docstring,
pyproject extra block, CLAUDE.md extras bullet, README install snippet.
Changes:
- Deleted `src/remove_ai_watermarks/face_restore.py` and its tests.
- Deleted the `restore` extra (gfpgan/facexlib/basicsr + scipy<1.18 / numba<0.60
pins) and the basicsr setuptools<69 build pin from pyproject.toml.
- Restored `src/remove_ai_watermarks/photomaker_restore.py` (V2 this time:
`TencentARC/PhotoMaker-V2`, `photomaker-v2.bin`, no `pm_version='v1'` override).
- Restored the `photomaker` extra in pyproject with all the upstream-compat
pins (einops, peft, onnxruntime, insightface) and the `allow-direct-references`
hatch metadata block.
- `InvisibleEngine` swapped `_restore_faces` -> `_restore_faces_photomaker`;
`--restore-faces-method` removed (only one method, no choice).
- CLI flag help, CLAUDE.md, README, docs/synthid.md, and
docs/controlnet-removal-pipeline-research.md all updated.
- docs/synthid-robust-identity-research.md status notice rewritten to list both
abandoned commercial-safe attempts (V1 + GFPGAN-on-cleaned) and the
non-commercial trade-off we accepted.
ruff + strict pyright(src/) clean; 578 tests pass (the 9 GFPGAN tests are gone,
the 11 PhotoMaker tests stay green).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
After 7 cascading upstream-compat fixes (insightface dep, peft dep, pm_version,
device, etc.), the PhotoMaker V1 cert sweep still hit a CFG batch-dim mismatch
inside the denoising loop. The upstream PhotoMaker `pipeline.py` is forked from
diffusers v0.29.1 and our env runs 0.38; SDXL prompt-encoder handling changed
significantly between those versions, so making PhotoMaker work end-to-end
needs a proper fork or a diffusers downgrade — both expensive. Not worth
shipping today.
Pivot: restore `face_restore.py` (GFPGAN) with a single-line fix that makes it
SynthID-safe by construction. The previous design ran GFPGAN.enhance on the
ORIGINAL watermarked image and was oracle-confirmed to re-add SynthID via the
weight-0.5 pixel blend. The fix is to run GFPGAN on the diffusion-CLEANED
image — whatever pixels GFPGAN derives from are already SynthID-free, so the
partial blend cannot transport the watermark. Identity fidelity is lower than
a true identity-as-embedding stack would deliver, but it ships and works.
Changes:
- `src/remove_ai_watermarks/face_restore.py` restored from pre-wipe state with
one line changed: `restorer.enhance(cleaned_bgr, ...)` instead of
`restorer.enhance(original_bgr, ...)`. `original_bgr` is kept as an unused
positional argument for API stability.
- `src/remove_ai_watermarks/photomaker_restore.py` and its tests REMOVED. The
research note (`docs/synthid-robust-identity-research.md`) keeps a "status
notice" documenting why PhotoMaker is parked for now and what the path back
in would look like.
- `pyproject.toml` `restore` extra restored (gfpgan/facexlib/basicsr +
scipy<1.18 + numba<0.60 pins + the basicsr setuptools<69 build pin), plus
`photomaker` extra (with its einops/insightface/peft pile) and the
`[tool.hatch.metadata] allow-direct-references = true` block REMOVED.
- `InvisibleEngine._restore_faces_photomaker` removed; `_restore_faces`
restored. The `--restore-faces` CLI flag and its plumbing through cmd_*
signatures are unchanged.
- CLAUDE.md, README.md, docs/synthid.md, docs/controlnet-removal-pipeline-
research.md updated to describe the shipped GFPGAN-on-cleaned design and to
reference PhotoMaker only as the parked alternative.
ruff + strict pyright(src/) clean; 578 tests pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The upstream PhotoMaker package's `__init__.py` unconditionally imports a
face-analyser class from its `insightface_package` submodule, so JUST importing
`PhotoMakerStableDiffusionXLPipeline` (the V1 pipeline class we use) raises
`ModuleNotFoundError: No module named 'insightface'` if insightface isn't
present in the env. The Modal cert sweep caught this on the V1 image.
Resolution: pin `insightface>=0.7.3` (and its `onnxruntime` runtime dep) in the
`photomaker` extra. The PyPI insightface package is MIT-licensed CODE; the
non-commercial restriction sits on the pretrained model packs (antelopev2,
buffalo_l) which download only when `FaceAnalysis()` is instantiated. Our V1 path
never instantiates the face-analyser -- it loads photomaker-v1.bin (CLIP-only
encoder) via `load_photomaker_adapter` -- so the model-pack license does not
bind us; we depend only on the MIT code for the import to resolve.
Safety guards:
- Runtime check in `_get_pipeline`: raises if `_PHOTOMAKER_FILE` is ever pointed
at v2 (so a future maintainer can't silently regress to the InsightFace path).
- New test class `TestV1OnlyCommercialSafetyGuard`: asserts repo + filename
pin to V1 AND asserts the module source never references the face-analyser
class (a static check that our codepath stays out of the runtime that would
pull the non-commercial model packs).
Docs: documented the import dance + legal split inline at the top of
`photomaker_restore.py`.
ruff clean; 581 tests pass (the 9 PhotoMaker tests plus 3 new V1-guard tests).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The GFPGAN `restore` extra and its `face_restore.py` module are gone. They were
oracle-confirmed to re-introduce SynthID by blending watermarked original face
pixels at fidelity weight 0.5 (clean A/B: gemini_3 controlnet 0.20 detected WITH
GFPGAN, clean WITHOUT). Keeping them as the default restore method was a footgun
for the removal pipeline. PhotoMaker-V2 (added in the previous commit) is the
single shipped restore path now -- identity-as-embedding, SynthID-safe by
construction.
Removed:
- src/remove_ai_watermarks/face_restore.py + tests/test_face_restore.py
- pyproject.toml `restore` extra (gfpgan/facexlib/basicsr + scipy/numba pins)
- pyproject.toml `[tool.uv.extra-build-dependencies] basicsr = [...]` build pin
- CLI: `--restore-faces-method` and `--restore-faces-weight` (no method choice
to make, no GFPGAN weight knob to expose)
- InvisibleEngine._restore_faces method (only _restore_faces_photomaker remains)
- All restore-faces-method / restore-faces-weight threading through cmd_*
signatures and _process_batch_image
Kept:
- `--restore-faces / --no-restore-faces`: now binds to PhotoMaker-V2.
- All adopted oracle findings about GFPGAN re-introducing SynthID (kept in the
research docs as historical context that explains why the path was removed).
Docs updated: CLAUDE.md (restore extras bullet collapsed to photomaker, removed
face_restore Key-modules bullet, several inline GFPGAN refs scrubbed), README.md
(face-identity callout + install section now point to the photomaker extra),
docs/synthid.md 5.5 (net recipe), docs/controlnet-removal-pipeline-research.md
(recommendations).
ruff + strict pyright (src/) clean; 578 tests pass (the 9 GFPGAN tests are gone,
the 9 PhotoMaker tests stay green).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds the second face-restore mechanism, selectable via the new CLI option
`--restore-faces-method=photomaker`. Unlike the existing GFPGAN path (which runs on
the watermarked ORIGINAL and was oracle-confirmed to re-introduce SynthID by partial
pixel blending), PhotoMaker carries identity in a SynthID-invariant OpenCLIP
embedding and regenerates fresh face pixels conditioned on it — the pixels in the
output are diffusion-fresh, so the watermark cannot be transported.
The load-bearing assumption (embedding invariance to SynthID-magnitude pixel noise)
was empirically validated in the prior commit (smoke test): cosine drift 0.002
under a ±2 LSB low-freq carrier, an order of magnitude less than JPEG90 drift
which SynthID survives at >=99% TPR.
End-to-end commercial-safe:
- PhotoMaker-V2 weights: Apache-2.0 (TencentARC)
- ID encoder: OpenCLIP-ViT-H/14 (MIT)
- SDXL base: shared with the main pipeline
- NO InsightFace (the non-commercial blocker for IP-Adapter FaceID / InstantID /
PuLID / Arc2Face)
Two-pass architecture (PhotoMaker has no ControlNetImg2img class in diffusers):
1) main controlnet/default removal pass cleans SynthID + drifts faces
2) PhotoMaker txt2img regenerates each face from its embedding, feather-composited
back into the cleaned image
New module `photomaker_restore.py` mirrors `face_restore.py`: lazy pipeline
singleton (double-checked lock), `is_available()` gate, pure `_face_crop_square` and
`_composite_faces` helpers, all unit-tested without the model (9 new tests). New
`InvisibleEngine._restore_faces_photomaker` runs after the diffusion pass, mirroring
`_restore_faces`. CLI flag `--restore-faces-method=[gfpgan|photomaker]` threaded
through `cmd_invisible`/`cmd_all`/`cmd_batch` + `_process_batch_image`.
New optional `photomaker` extra (Apache-2.0 + Apache-2.0/MIT deps, no basicsr).
`[tool.hatch.metadata] allow-direct-references = true` is required because the
upstream PhotoMaker package lives only on GitHub.
The next step (separate work) is oracle validation: run a 6-image cert sweep
through the new pipeline (default/controlnet at the certified strength +
--restore-faces-method=photomaker) and confirm SynthID stays clean while face
identity is recovered. The required infrastructure (`raiw-app/modal_cert.py`) is
already in place.
ruff + strict pyright(src/) clean; 586 tests pass (+ 9 new in
tests/test_photomaker_restore.py).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
New samsung_engine.py mirrors the jimeng engine but anchors bottom-left; wired
into watermark_registry, the CLI (--mark samsung / auto), and identify
(visible_samsung, medium). visible_alpha_solve.py gains a corner=bl mode;
samsung_alpha.png solved from @f-liva's flat captures. Calibrated for the
Italian "Contenuti generati dall'AI" variant. Flat black/gray/white captures
committed, real photos gitignored. Tests + docs.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The fp16-fix VAE swap (#29) is gated to the default SDXL checkpoint, so a
custom model_id, a stale pre-fix install, or a fal/custom loader can still
decode to an all-black/NaN frame in fp16 (reporter: gpt-image 1448x1086,
the `image_processor.py invalid value encountered in cast` warning).
Add a model-agnostic backstop in remove_watermark: after generation, if the
run was fp16 and the output is degenerate (_is_degenerate_image: near-zero
mean and variance), rebuild the pipeline in fp32 on the same device and
re-run once. fp32 is the verified-clean path, so a black image is never
returned regardless of model_id or version. Mirrors the MPS->CPU fallback's
self-mutation pattern; batch inherits it. Verified e2e on MPS by forcing
fp16 with the swap disabled (first pass black, guard fired, retry clean).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
After reverse-alpha, re-detect the sparkle; when one survives at or above the
registry fail line (conf >= 0.5) -- an alpha mismatch the per-image gain estimate
could not fully correct -- inpaint the footprint and keep that only when it lowers
the re-detect confidence. The footprint inpaint reconstructs the slot from its
darker surroundings, so it physically removes the bright sparkle; purely additive,
the common clean removal re-detects below 0.5 and is returned untouched.
Measured on the spaces visible-removal audit: gemini removal-audit failures drop
15 -> 11 (4 genuine rescues), doubao 65/65 and jimeng 11/11 unchanged, zero
regressions on the 468 already-clean removals.
An offset+scale alignment search was prototyped on the remaining 11 fails and
rejected: an audit "ceiling" suggested +4 more, but those were NCC-gaming -- the
lower-scoring placement left the sparkle as bright or brighter, just reshaping the
residual so the contrast-invariant shape-NCC scored lower (a5a9: first-pass slot
~76 at background level vs the "aligned win" ~164). A brightness sanity check
rejected every one, so it contributed nothing and was removed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three content-quality features for the invisible/all/batch pipeline.
DBNet text detector (auto_config): replace the MSER text heuristic with
PP-OCRv3 differentiable-binarization via cv2.dnn.TextDetectionModel_DB,
using a bundled 2.4 MB Apache-2.0 model (en/cn detection nets are
byte-identical, so it ships language-neutral). cv2.dnn is core OpenCV, so
no new pip dep. MSER stays as the fallback when the model can't load.
Validated on real images: matches MSER everywhere and additionally catches
the Doubao CJK mark MSER missed; routing decisions unchanged otherwise.
Real-ESRGAN upscaler (new upscaler.py, esrgan extra): optional
pre-diffusion super-resolution for the min-resolution floor upscale, loaded
via spandrel (MIT, no basicsr) with BSD-3-Clause weights downloaded on
first use. New --upscaler {lanczos,esrgan} on invisible/all/batch; default
stays lanczos and the engine falls back to lanczos when the extra is absent
or the model errors (never breaks removal). It is a manual opt-in knob (the
auto plan never selects it) -- as a generic GAN it sharpens photo/texture
content strongly but can degrade faces (the diffusion pass regenerates
them) and thin text, documented accordingly.
batch --auto: wire the content-adaptive --auto (+ --adaptive-polish) into
cmd_batch. The plan is recomputed per image and the invisible engine is
cached per resolved pipeline (default/controlnet), so a mixed directory
builds at most one engine of each kind. Verified end-to-end: 3 mixed
images routed correctly with only 2 pipeline loads (controlnet reused).
ruff + strict pyright(src/) clean; 558 tests pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Brings in commit 5cf68a6 (single C2PA_AI_VENDORS registry, erase_lama
grayscale/BGRA support, batch device-cache clearing + --controlnet-scale,
uv publish via OIDC, hatchling pin <1.31). Auto-merged with no conflicts;
ruff/pytest(544)/pyright all clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
detect_watermark's shape-only NCC (spatial/gradient/var fusion) fires on ornate
or flat content (text strips, banners, hatching) that coincidentally matches the
diamond shape. The NCC is contrast-invariant, so it cannot see the defining
property of a real Gemini sparkle: a bright WHITE overlay whose core sits above
the local background.
The fusion now demotes (caps confidence to 0.30) a match that is BOTH
low-confidence (< _SPARKLE_FP_CONF 0.65) AND has a low core-ring brightness
margin (_core_ring_margin < _SPARKLE_FP_MARGIN 5). Real sparkles escape via
EITHER high confidence (white-bg sparkles score >=0.79 despite a low margin) OR
high margin (dark/mid backgrounds, incl. the #36 faint-corner case), so both
must fail to demote. The gate is monotonic -- it only removes detections, never
adds -- so it cannot regress the verified-negative corpus (already 0 FPs).
On the spaces corpus it demoted 16/495 flagged sparkles (13 no AI metadata =
content FPs; the 3 AI-meta ones were visually FPs / a near-invisible
white-on-white sparkle whose AI verdict is held by metadata), and dropped the
removal-audit failures 20 -> 15.
- _core_and_bg shared helper (core 75th-pct brightness vs background-ring median);
_estimate_alpha_gain refactored onto it, new _core_ring_margin wrapper.
- TestSparkleFalsePositiveGate: margin high/low, strong-sparkle kept (incl. on
white via high conf), blurred no-core blob demoted.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The fixed mild auto polish (unsharp 0.5 / grain 2.0) under-corrected soft
photo/face output (gemini_3 stayed at lap-var 84 vs its 592 original) and its
grain speckled small text. Replace it with humanizer.adaptive_polish: target the
input's Laplacian variance with a capped unsharp scaled to the deficit + edge-
masked grain (smooth regions only), calibrated by a short sigma search. Self-
limiting on text/graphics -- already high-frequency, so almost no polish lands
and text edges are masked out. Validated on the spaces corpus (gemini_3 84 -> 334
end-to-end; openai_1 text near-untouched).
Interface: every --auto decision is now independently overridable -- add
--adaptive-polish/--no-adaptive-polish (matching --restore-faces; works without
--auto too) so the polish can be disabled or used manually. _apply_auto overrides
exactly the three content-adaptive modes (pipeline, restore-faces, adaptive-
polish); --unsharp/--humanize stay independent fixed filters.
cv2-only, no new deps. Threaded through invisible/all (not batch).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Three P2 cleanups from a library-wide review.
Detection -- single C2PA_AI_VENDORS registry (noai/constants.py):
- C2PA_ISSUERS, SYNTHID_C2PA_ISSUERS, and identify._ISSUER_PLATFORM now derive
from one C2paAiVendor table, so adding a C2PA vendor is one entry instead of
edits in three places across two files. Behavior-identical (262 detection
tests pass; the kept `needle` field is load-bearing -- it differs from `org`
for Google and ByteDance, with no mechanical derivation).
Code-health:
- region_eraser.erase_lama now accepts grayscale/BGRA like erase_cv2 (it
crashed on grayscale and silently dropped alpha on BGRA). +2 regression tests.
- batch frees the device cache between images via a shared try_empty_device_cache
helper (generalized from the MPS-only _try_clear_mps_cache, now reused by both
the MPS->CPU fallback and the batch loop).
- batch gained --controlnet-scale (parity with invisible/all).
CI / packaging:
- publish.yml uploads via `uv publish` (PyPI trusted publishing over OIDC),
replacing pypa/gh-action-pypi-publish so uploads no longer depend on that
action's bundled twine accepting the Metadata-Version. Workflow filename +
pypi environment unchanged, so PyPI's trusted-publisher entry still matches.
- hatchling pin relaxed <1.28 -> <1.31 (verified against hatch's changelog:
1.30.0 made Metadata 2.5 the default, 1.30.1 reverted to 2.4; 1.27-1.29 were
always 2.4). Kept as belt-and-suspenders so the first uv-publish release ships
2.4, isolating the uploader swap from the metadata-version bump.
Docs (CLAUDE.md, pyproject) synced; corrected the inaccurate "hatchling 1.28+
emits 2.5" note.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add `auto_config.plan(image_path) -> AutoConfig`, the first step of the
invisible/all pipeline: it inspects the input image (before the diffusion model
loads) and picks the quality modes so the run adapts to content. Quality-priority
routing -- ControlNet (text/face-structure preservation) is the default, skipped for
plain SDXL only on a clearly structure-less image; GFPGAN face restore when a face is
present; a mild sharpen + grain polish when a smoothing pass ran. Exposed as `--auto`
on `all`/`invisible` (`_apply_auto`; explicit flags override via click's parameter
source). Not wired into batch (its engine is cached per-mode).
Detection is cv2-only and torch-free (~100 MB peak RSS, a few ms): OpenCV YuNet
(`cv2.FaceDetectorYN`, MIT, 232 KB model bundled in assets/) for faces, a Canny
edge-density + MSER heuristic for text/structure (a rough Phase-1 placeholder; DBNet
via cv2.dnn is the planned upgrade). ZERO new pip deps. Designed to run wherever the
pipeline runs -- the raiw.cc Modal GPU worker -- never on the 512 MB web host.
Real-ESRGAN-via-Spandrel upscaling (a new `esrgan` extra) and an adaptive
Laplacian-variance polish are deferred to later phases.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The captured sparkle alpha peaks ~0.51, but some real Gemini sparkles are
rendered more opaque. The fixed-alpha reverse blend then UNDER-subtracts and
leaves a bright residual the detector still fires on. A visible-removal audit
through the registry path on the spaces corpus showed this as a meaningful
fraction of marks -- all under-removals, not a background-brightness class
(failures and successes had the same input confidence and background luma; the
discriminator was the removal delta itself).
remove_watermark now estimates a per-image alpha gain (_estimate_alpha_gain:
effective sparkle opacity at the bright core vs the local background ring,
a_eff/a_cap, clamped [1.0, 1.94]) and scales the alpha to match before the
over-sub/blend branch. A 1.05 deadband keeps a sparkle that already matches the
capture byte-identical to the pre-fix output, so the fix is purely additive
(0 regressions on the audit set; failures dropped substantially). The over-sub
guard still runs on the scaled alpha as the safety net for an over-shoot.
- _estimate_alpha_gain + _ALPHA_GAIN_MAX/_DEADBAND/_CORE_FRAC in gemini_engine.
- TestUnderSubtractionGain asserts on footprint pixels, NOT the detector (its
NCC is degenerate on a flat synthetic bg; the real corpus removal drops the
detector ~0.80 -> ~0.27).
- scripts/visible_removal_audit.py: the detect -> remove -> re-detect audit tool
that found and validated this (operates on gitignored data/spaces only).
- CLAUDE.md + README: document the under-subtraction gain.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two quality knobs for the SDXL invisible pass:
- min_resolution floor (default 1024, --min-resolution): small inputs are
upscaled to a 1024px long-side floor before diffusion, since SDXL img2img
distorts on a tiny latent (a 381x512 portrait wrecks at native). The output
is restored to the original input size, so it is a transparent quality boost;
it adds time/memory on small inputs. 0 disables. Extends the pure _target_size
helper (now cap-or-floor-or-native, min skipped on a min>max misconfig),
unit-tested without a model.
- unsharp post-filter (humanizer.unsharp_mask, --unsharp, opt-in default 0):
applied LAST, after the GFPGAN face pass (a pre-GFPGAN sharpen would be
smoothed back over), to counter the soft/over-smoothed look that diffusion +
restoration leave behind (an AI tell). Pairs with --humanize (grain).
Both threaded through invisible/all/batch + the module-level helper. Verified
end-to-end on a 381x512 portrait: upscaled to 1024, sharpened, restored to
381x512.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add an optional, commercial-safe face-restoration post-pass that recovers
face identity the diffusion removal pass drifts (canny holds structure, not
likeness) while still scrubbing the pixel watermark in the face regions.
- face_restore.py: GFPGANer singleton (CPU unless CUDA), the basicsr
torchvision.transforms.functional_tensor shim, and the pure feather
_composite_faces helper (unit-tested without the model). GFPGAN
re-synthesizes each face from a StyleGAN2 prior, so composited face pixels
are GAN-generated (no watermark, no pixel-copy) -- oracle-clean at weight 0.5
with identity preserved.
- InvisibleEngine.remove_watermark: restore_faces / restore_faces_weight,
best-effort, auto-skips when the extra is absent or no face is detected.
- CLI --restore-faces/--no-restore-faces + --restore-faces-weight on
invisible/all/batch (on by default).
- restore extra (gfpgan/facexlib/basicsr), numpy<2-pinned (scipy<1.18,
numba<0.60) and kept out of `all`; basicsr needs Python <3.13 + setuptools<69
to build, so pin .python-version 3.12.
Commercial-safe: GFPGAN Apache-2.0, RetinaFace MIT. The CodeFormer alternative
is non-commercial and is not shipped. The earlier IP-Adapter FaceID layer was
removed (footgun: needs high strength, corrupts faces at the low removal
strength).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add `--pipeline controlnet` (SDXL base + xinsir canny ControlNet via
StableDiffusionXLControlNetImg2ImgPipeline): the canny edge map conditions the
img2img regeneration so text and face STRUCTURE stay sharp, while the watermark
is still removed by the regeneration (`strength`) -- no original pixels are
copied or frozen, so SynthID does not survive. Oracle-verified clean on OpenAI
with better text/structure fidelity than plain img2img at equal strength.
`--controlnet-scale` tunes structure preservation; fp32 on mps/cpu (fp16-fixed
VAE on cuda/xpu). Shares the img2img runner (live progress + MPS->CPU fallback)
and the fp16-VAE-fix / device-move helpers with the default pipeline.
Remove the superseded subsystems -- ctrlregen (SD1.5 clean-noise),
text-protection (differential / region-hires) and face-protection: they either
destroyed real content or shielded the watermark by re-using original pixels.
controlnet replaces them by regenerating everything under edge conditioning.
Canny preserves face structure but not identity; face IDENTITY is a separate
face-restoration post-pass (CodeFormer/GFPGAN), researched + prototyped but not
yet shipped. An IP-Adapter FaceID attempt was built and removed (footgun: needs
high strength, corrupts faces at removal strength).
Docs: docs/controlnet-removal-pipeline-research.md, scripts/controlnet_sweep.py.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
detect_watermark's size-weighted global NCC search lets a larger, mediocre
match (e.g. a bright collar in a portrait) outrank a small, near-perfect
sparkle in the bottom-right corner, so a faint sparkle on a busy background
scored below threshold and the image read as clean -- the regression from
widening the search window 256px->512px between v0.7.2 and v0.8.8.
Add _corner_promote: a bottom-right-corner raw-NCC pass that overrides the
global pick when the corner holds a match with raw NCC >= 0.85 that beats it.
It only ever replaces a lower-fidelity pick (cannot weaken an existing
detection) and keeps the wider window for variant margins. The corner side is
relative-clamped (0.20 of the short side, [96, 384]) so it stays a true corner
at every scale: a fixed 256px covers ~70% of a small portrait, where a real
photo raw-matches the star at ~0.81; relative tightening drops that to ~0.69.
The 0.85 gate sits between the worst real-photo corner match (~0.78) and a
genuine faint sparkle (~0.93): zero false positives across native + downscaled
negatives, headshot rescued from below-threshold to 0.71.
Factor the shared multi-scale matchTemplate loop into _scan_scales.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The C2PA issuer attribution (`c2pa`) and the SynthID proxy (`synthid`) are
derived from the same manifest, so treating them as independent signals made
rule 1 fire on legitimate multi-actor manifests where a product wraps another
vendor's engine (Microsoft Designer on OpenAI, Microsoft on Google) or an edit
chain re-signs (Adobe over a Gemini original). 19 such files in the
2026-06-01/02 spaces batches read as "likely spoofed/laundered" before this.
Group `c2pa` + `synthid` into one provenance source via `_CLASH_SOURCE`; rule 1
now requires two vendors from different sources. A manifest vendor still clashes
with a genuinely independent stamp (EXIF/XMP generator, IPTC AISystemUsed, AIGC,
xAI).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
On a dark/textured background (e.g. grass) the captured alpha map over-estimates
the real Gemini sparkle's effective opacity (~0.51 captured vs ~0.31 effective),
so the fixed-alpha reverse blend over-subtracts (watermarked - alpha*logo goes
negative) and drives the footprint to black -- the white sparkle turns into a
black diamond (issue #30, reported by @CoolZimo1).
remove_watermark now detects this via _reverse_alpha_oversubtracts (fraction of
footprint pixels with a negative numerator > 5%) and inpaints the small sparkle
footprint from the surrounding pixels (cv2 NS, cropped to a padded box) instead.
Behavior-neutral on the working case: a bright background over-subtracts at ~0%,
so reverse-alpha is used and the output is byte-identical to before (verified:
demo_banana 0.0 frac vs the issue-#30 grass image 0.61 frac; issue-#30 footprint
recovers to background grass with no pit, residual sparkle conf 0.25 < 0.35).
Guard is scoped to GeminiEngine: doubao/jimeng already NCC-align their alpha to
the actual mark per image, which sidesteps the fixed-alpha mismatch.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The default img2img strength is now chosen from the detected SynthID vendor
(C2PA issuer) instead of a single fixed 0.30: OpenAI gpt-image -> 0.10, Google
Gemini -> 0.15, unknown source -> 0.15. Explicit --strength always wins.
Basis: an oracle-verified June 2026 controlled study (clean v0.8.6, text/face
protection OFF, per-image openai.com/verify or Gemini-app verdict). OpenAI's
SynthID clears at 0.05 across 1024-1600 px (n=4, resolution-independent);
Google's is ~3x more robust and needs 0.15 on the capped-1536 path (n=4). The
dominant factor is the VENDOR, not resolution. The earlier single 0.30 default
and the "resolution dependence" lore came from contaminated tests run with the
protect-text bug ON (issue #14) -- re-running those same 1600x1600 images clean
removes SynthID at 0.05.
`vendor_for_strength(path)` reads metadata.synthid_source on the ORIGINAL input
and is threaded through cli (invisible/all/batch) -> invisible_engine ->
watermark_remover -> resolve_strength(strength, profile, vendor), so display and
execution use the same vendor (the engine sees a temp path whose C2PA the visible
pass already stripped, so detection must happen in the CLI on the pristine
source). Caveat: Google's 0.15 was validated only on --max-resolution 1536;
native 2816 Gemini was not locally measurable (OOM on Apple Silicon) and is
pending GPU validation on raiw.cc.
Docs: docs/synthid.md sections 2.2/4.4/5.2 corrected (the contaminated
resolution-dependence findings replaced with the clean oracle-verified table);
README and CLAUDE.md updated; CLI --strength help reflects the adaptive default.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
identify previously ran only the Gemini sparkle as a visible detector, so a
Doubao/Jimeng image with stripped TC260 metadata had no visible fallback. Add
`_visible_text_marks` (registry-backed) so the ByteDance Doubao 豆包AI生成 and
Jimeng 即梦AI marks are detected too, each gated by its own engine NCC threshold
via MarkDetection.detected. New signals `visible_doubao` / `visible_jimeng`
(medium), same stripped-metadata fallback role as the sparkle; excluded from
integrity-clash vendor claims; set platform only when no harder signal did.
Also make `noai/__init__` lazy (PEP 562 __getattr__): importing the light
`noai.c2pa` / `noai.constants` submodules (which identify needs) no longer
eagerly pulls `watermark_remover`, which imports torch + diffusers at module
top. `import remove_ai_watermarks.identify` drops from ~420 MB to ~21 MB in a
full gpu/detect install (torch not loaded), so it fits a 512 MB host; the
removal API resolves lazily on first access. Guarded by TestIdentifyImportIsLight.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
An oracle-verified GPU strength study (Modal A100, native res, Gemini-app
'Verify with SynthID', n=3 fresh Gemini images, protect_text/faces off) found the
current Google SynthID survives strength 0.10/0.15/0.2 and is removed only at 0.3.
The previous 0.10 default (set from an n=1 result) no longer clears it -- Google
hardened SynthID and the threshold has climbed 0.05 -> 0.10 -> ~0.3. Bump
DEFAULT_STRENGTH to 0.30; OpenAI/ChatGPT carry C2PA not SynthID, so 0.10 is plenty
there (pass --strength 0.10). Note protect_text shields the text regions SynthID
hides in (use --no-protect-text for full removal on text-heavy images).
The same study found ctrlregen at clean-noise strength DESTROYS real images
(hallucinated micro-text in smooth regions), with no usable middle setting, so the
literature's 'clean-noise is the lever' did not hold empirically. Flag ctrlregen
EXPERIMENTAL in the CLI --pipeline help, README, and watermark_profiles; SDXL
img2img at ~0.3 stays the shippable path.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
cli.py now emits plain ASCII through a small click.echo shim
(_Console / _Table / _Progress) instead of rich: no colors, markup tags,
panels, progress bar, or Unicode glyphs (Warning: / -> / ... and dropped
checkmark/cross marks). identify and metadata tables render as indented
plain lines.
- drop rich from dependencies (pyproject.toml + uv.lock)
- __init__: set TRANSFORMERS_VERBOSITY=error (setdefault) plus a warnings
filter so the transformers Siglip2ImageProcessorFast deprecation no
longer prints at CLI startup (it fires from the eager noai import)
- TestGpuHintMarkup: the [gpu] hint is now printed verbatim; docstring updated
- CLAUDE.md: replace the obsolete rich-markup lesson, note the verbosity fix
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The ctrlregen profile inherited the SDXL img2img --strength default (0.10), a
near-identity pass that loaded ControlNet + DINOv2-giant and barely changed the
image -- a no-op for removal. resolve_strength() now resolves an unset strength
per profile: 0.10 for the SDXL default, CTRLREGEN_DEFAULT_STRENGTH (1.0,
clean-noise) for ctrlregen. It checks `is None` rather than falsiness, so an
explicit 0.0 is respected (the old `strength or DEFAULT` swallowed it).
Research basis: CtrlRegen (ICLR 2025, arXiv:2410.05470) removes robust
watermarks by regenerating from clean Gaussian noise; partial-noise img2img
retains watermark info that diffuses back, so a high (clean-noise) strength is
the lever, not a knob on the light SDXL pass. CLI wiring (--strength default
None) lands with the cli refactor.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Visible-watermark work across all three corner-mark engines plus a committed,
reproducible alpha-build pipeline (scripts/visible_alpha_solve.py) fed by committed
solid black/gray/white captures.
- jimeng: new "即梦AI" wordmark remover (reverse-alpha + thin residual inpaint,
always NCC-aligned -- the mark re-rasterizes/jitters per image). Detect via glyph
silhouette NCC (0.45 threshold; does not cross-fire with Doubao). Registered in the
visible-mark catalog; `visible --mark jimeng` / `--mark auto`.
- doubao: fix a real production defect -- the shipped remover left a READABLE
"豆包AI生成" outline on real samples while detect() returned conf 0.0 (fooled by a
thin outline), so the test passed and the "56/56 clean" claim was detector-measured,
not visual. Root cause: under-estimated alpha + fixed-geometry-no-inpaint + tight
locate box. Rebuilt alpha (careful gray-self solve), always-align, thin inpaint,
widened locate box -> readable outline becomes faint texture-level traces.
- gemini: rebuild gemini_bg_{96,48} from our own controlled captures (validated NCC
0.9998 vs the prior third-party asset); removal re-verified clean, no behaviour change.
- tests: add textured-shift regression to both engines (guards the align-on-shift path
the Doubao defect exposed; lesson: a detector-only removal test is insufficient,
assert visual residual).
- docs: CLAUDE.md, README, capture READMEs and docstrings synced; stale
"exact/pixel-exact/56-clean" claims removed.
Also includes a SynthID label-wording clarification in identify.py/cli.py
("SynthID pixel watermark" -> "SynthID watermark, inferred from C2PA metadata").
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Apply fixes from a full-repo review (code, tests, docs).
Security / correctness:
- Clamp attacker-controlled PNG/caBX chunk lengths to the remaining file
size in metadata.py and noai/c2pa.py (a malformed length no longer drives
a multi-GB read); skipped chunks seek instead of read.
- noai/isobmff.strip_c2pa_boxes is now fail-safe on a malformed box: return
the original bytes with a warning instead of silently truncating the tail,
so metadata --remove can no longer emit a corrupt file.
- doubao_engine._fixed_alpha_map clamps the glyph box to the image (no crash
on degenerate width-vs-height).
- watermark_remover._run_region_hires gates the phaseCorrelate offset on
response and magnitude (a spurious shift no longer garbles text) and drops
the generator after a CPU fallback (no MPS/CPU device mismatch).
Robustness:
- gemini_engine, doubao_engine, region_eraser normalize grayscale and RGBA
inputs to BGR at the engine entry points.
- image_io.imwrite returns False on an unwritable path (matches cv2).
- invisible_engine guards a None imread result before use.
- trustmark_detector._decoder uses a double-checked threading lock.
- ctrlregen.tiling.tile_positions raises on overlap >= tile.
- humanizer chromatic shift no longer wraps opposite-edge pixels.
- identify OpenAI caveat keyed on the normalized vendor, not a substring.
- Remove the dead "visible --detect-threshold" CLI option.
- publish.yml verifies the release tag matches the package version.
Docs:
- README strength 0.05 to 0.10; .env.example HF_TOKEN marked optional;
doubao_capture README updated to reverse-alpha-only; CLAUDE.md synced with
the new behaviors and the batch command.
Tests: new test_security_clamp.py for the read clamp and isobmff fail-safe;
erase CLI coverage; integrity-clash rule 2 end-to-end; multi-tag EXIF
survival and cross-format strip guards; channel/size, tiling, humanizer, and
imwrite regressions. Full suite 493 passed, 2 skipped; ruff and pyright src/
clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The stock SDXL VAE overflows to NaN in fp16, so the plain img2img path decodes
to an all-black image on a CUDA/XPU fp16 backend. This is the raiw.cc black
result HitaoLin reported (a 1086x1448 input came back uniformly black). cpu/mps
run fp32 and never hit it, and the differential / region-hires pipeline already
upcasts the VAE itself, so only the plain path on a fp16 GPU was exposed.
`_load_pipeline` now loads `madebyollin/sdxl-vae-fp16-fix` for the default SDXL
checkpoint when running fp16, gated by the pure helper `_needs_fp16_vae_fix`. A
custom non-SDXL model keeps its own VAE.
The decision logic is unit-tested without a download (TestFp16VaeFix). The
black->clean recovery itself needs a CUDA GPU and was not verifiable on this MPS
machine; it must be confirmed on the backend.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* Raise default SynthID-removal strength 0.05 -> 0.10 (current Google SynthID)
The old default (0.04/0.05) no longer removes the CURRENT Google SynthID (Nano
Banana / Gemini 3): verified 2026-05-30 via the Gemini 'Verify with SynthID'
oracle on a real image -- 0.05 still detected, 0.10 not detected (OpenAI's was
already cleared at 0.05). Add DEFAULT_STRENGTH=0.10 in watermark_profiles, route
the engine + CLI defaults to it. At 0.10 small text deforms more, which is why
text protection (_run_region_hires) runs by default. CLAUDE.md SynthID note
corrected. CAVEAT: n=1 Google + n=1 OpenAI; broad corpus oracle validation
pending (task tracked).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* Drop unused LOW/MEDIUM/HIGH strength profiles; CLI --strength defaults to DEFAULT_STRENGTH
The fixed strength presets (and get_recommended_strength) were dead -- nothing in
the pipeline used them, only tests. One knob now: DEFAULT_STRENGTH (0.10),
overridable per-call via the CLI --strength flag, which now defaults to that
constant (single source of truth). Removed the WatermarkRemover.LOW/MEDIUM/HIGH
class attrs and the get_recommended_strength tests.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace the default text-protection path. Differential Diffusion froze text in
latent space, which left SynthID intact inside text (violating remove-everywhere)
and still softened sub-8px strokes (VAE latent limit). _run_region_hires instead
scrubs the whole image, then re-scrubs each detected text block at high resolution
and feather-composites it back: every pixel is regenerated (watermark removed
everywhere) while small text stays crisp (high-res strokes span >1 latent cell).
merge_text_regions + feather_paste are pure and unit-tested; each re-scrubbed
patch is phase-correlated back to the original crop to null the ~1-2px round-trip
offset. Synthetic 18px multilingual text: text-region SSIM 0.28 -> 0.48, visually
garbled -> readable across Latin/Cyrillic/CJK. Legacy _run_differential /
build_change_map remain but are no longer the default. Prod use still requires
confirming via the SynthID oracle that re-scrubbed text zones read watermark-free.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
On RGBA inputs the CLI forced the watermark bbox alpha to 0 on save, so the
removed-sparkle area became a transparent hole that renders as a solid white
box on any non-transparent viewer. The Gemini app exports opaque RGBA, so
every user hit it. Reverse-alpha already recovers the real pixels there (and
`erase` inpaints them), so there is no artifact to hide -- the hole was the
bug, introduced as an over-correction in d091b9f.
`_write_bgr_with_alpha` now rejoins the input alpha plane unchanged (drops the
`clear_region`/`pad` params); the `visible` / `erase` / `all` / `batch` call
sites drop the cleared-region argument and the orphaned region bookkeeping.
The registry `remove()` still returns the mark bbox (used for inpaint_residual
positioning); the CLI just no longer clears alpha with it.
Inverts the test that locked in the old behavior into a #30 regression guard
(watermark-region alpha stays opaque, no pixel forced transparent). Verified
end-to-end on a real Gemini RGBA export: sparkle gone, zero transparent
pixels, clean over a white background.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A corpus audit surfaced China TC260 AIGC-labeled images that `identify`
missed. Three detection gaps in `aigc_label`, all fixed:
- raw-JSON `{"AIGC":{...}}` in JPEG EXIF (UserComment): brace-matched from
the scan head with `json.raw_decode`, gated on a TC260 field like the
PNG-chunk path. (Doubao-class output via that export surface.)
- XMP attribute form `TC260:AIGC="{...}"` (PicWish): folded into the
element regex as a second alternation.
- TC260 XMP packet appended after a large `IDAT`, past the 1 MB scan
window: `scan_head` now appends late PNG metadata chunks via
`_png_late_metadata`, mirroring the existing ISOBMFF late-box scan.
Adds `scripts/corpus_gap_scan.py`: runs `identify` over a corpus, writes
the per-file report CSV, and flags `unknown` files that carry a known
marker in their metadata region (the audit that found these gaps).
Scanning only the metadata region — not the whole file — avoids the
random short-token collisions inside compressed PNG/JPEG streams.
On the local corpus this lifts 3 files from `unknown` to AI (China AIGC)
and leaves zero false gap candidates. Synthetic piexif/PngInfo fixtures
cover all three forms.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* fix(trustmark): gate detection on re-encode durability to kill false positives
TrustMark's wm_present flag is a BCH validity check that spuriously
validates on a content-correlated fraction of un-watermarked images
(AI textures trip it more than camera photos). On a 1343-image set all
20 raw detections were false, several on Gemini/OpenAI/Doubao output that
cannot carry Adobe's watermark, with random-bytes secrets.
A genuine TrustMark is a durable soft binding that survives re-encoding,
so detect_trustmark now re-decodes after a mild JPEG round-trip and
requires the same schema both times. Every observed false positive
collapsed under this gate; the second decode runs only on the rare hit.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(identify): Samsung Galaxy AI, FLUX, ByteDance C2PA; fix C2PA substring FP
Detection extensions verified on real signed files (2026-05-29):
- Samsung Galaxy AI: signer attribution via a new _SIGNER_C2PA_PLATFORM
(Samsung Galaxy / ASUS Gallery) kept separate from the capture-camera
_DEVICE_C2PA_PLATFORM so a Galaxy AI edit (device cert + AI source type)
does not trip the camera-vs-AI integrity clash. Plus metadata.samsung_genai:
the proprietary genAIType marker in PhotoEditor_Re_Edit_Data, a medium-
confidence AI-editing signal (samsung_only branch).
- Black Forest Labs (FLUX) and ByteDance Volcano Engine (Doubao/Jimeng)
added as C2PA issuers + issuer->platform mappings.
- fix: C2PA presence required only the bare 4-byte 'c2pa' substring, which
false-positives on compressed pixel data (a recompressed PNG IDAT re-flagged
C2PA after its manifest was correctly stripped). New c2pa_marker_in() requires
the JUMBF wrapper (jumb+c2pa) or the C2PA uuid box; applied in identify +
metadata. Verified: all 535 real C2PA files carry jumb.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(doubao): gate detection on text structure to cut ~95% of false positives (#23)
Coverage alone over-fired: any textured bottom-right corner cleared the
threshold, so the detector false-positived on ~28% of arbitrary images.
The real '豆包AI生成' mark is six glyphs in one row, so detect now also
requires the text-structure signature (_glyph_structure): many connected
components, no single dominant blob, concentration in a thin horizontal
band. False positives dropped 343 -> 17 across the corpus while keeping
real-mark recall and the doubao-1.png sample. Also accept a no-op force
kwarg for remover-interface symmetry.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(samsung): add Samsung Galaxy AI visible-badge remover
New samsung_engine.py removes the bottom-left sparkle + localized
'AI-generated content' badge that Galaxy AI tools stamp. Mirrors the
Doubao locate->mask->inpaint pattern but bottom-left, with a dual-polarity
top-hat mask (the badge is light-on-dark or dark-on-light). Detection gates
on a band + left-anchor signature (the Doubao CJK-component gate does not
transfer: Latin badge letters connect into few blobs). Explicit-only --
tuned on few real badges with a ~4% FP floor, so it is not used in auto.
Synthetic byte-blob fixtures (real badges are user content, not shipped).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(visible): unified known-watermark registry + LaMa inpaint backend
watermark_registry.py is a single catalog of known visible marks, each
tying {usual location, in_auto flag, recovery strategy, detect adapter,
remove adapter}: gemini (reverse-alpha, exact), doubao, samsung. cmd_visible
is now registry-driven (best_auto_mark for --mark auto; mark_keys() feeds the
CLI choices) -- the per-mark _run_doubao/_run_samsung helper branches are gone.
Cross-engine confidences are not comparable, so the gemini adapter applies the
corpus-validated 0.5 sparkle threshold for auto arbitration (its engine flag is
loose and weakly fired ~0.36 on Doubao text, hijacking auto).
--backend auto|cv2|lama chooses background reconstruction for the mask-based
marks; auto = LaMa when onnxruntime is present, else cv2. For LaMa the mask is
the FILLED glyph bounding box (sparse glyph masks leave anti-aliased edges
behind). cv2 stays the zero-dependency fallback.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* docs: watermark registry, Samsung/FLUX/ByteDance detection, LaMa backend, trustmark gate
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(doubao): exact reverse-alpha removal from captured alpha map
The Doubao '豆包AI生成' mark is a fixed semi-transparent white overlay, so
given its alpha map the original pixels are recovered exactly:
original = (wm - a*logo)/(1-a) -- no inpaint hallucination.
The alpha map + logo colour were solved from real black+gray Doubao captures
on a controlled background: on black captured = a*logo, and the black/gray pair
solves a per-pixel without assuming the logo colour (a_max~0.65, logo near-white);
the white capture cross-validates (mark vanishes to a flat fill). Bundled as
assets/doubao_alpha.png + geometry constants.
remove_watermark_reverse_alpha applies it scaled to image width; exact at the
captured width, so the registry routes doubao through it only when
reverse_alpha_available (width within the calibrated band) and the mark is
detected, falling back to mask inpaint (cv2/LaMa) otherwise. A light residual
inpaint cleans the sub-pixel rescaling error. Add captures at more resolutions
to widen exact coverage.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* refactor(visible): reverse-alpha only -- drop inpaint removal + heuristic detection
Per the principle that we only remove/detect what we can do exactly, the
visible-mark path is now reverse-alpha only:
- Doubao detect is reverse-alpha-consistent: match the bundled alpha glyph
silhouette against the corner via TM_CCOEFF_NORMED (DETECT_NCC_THRESHOLD 0.4)
-- keys on the '豆包AI生成' SHAPE, not coverage/structure heuristics. FP
7/1243 (0.6%). Removes the cv2 inpaint path + the _glyph_structure gate.
- Registry is reverse-alpha only: dropped the cv2/LaMa backend (_glyph_remove,
_lama_box_inpaint, default_backend, --backend) and the Samsung entry. Doubao
outside the alpha resolution band is skipped, never inpainted.
- Removed samsung_engine.py + tests + --mark samsung (no alpha map captured;
Samsung C2PA/genAIType metadata detection in identify is unaffected).
- The universal erase --region (cv2/LaMa) is unchanged -- arbitrary-region
inpainting stays a user-directed tool, separate from the known-mark registry.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(doubao): NCC sub-pixel alignment -> reverse-alpha at any resolution
A pure width-scale of the captured alpha map is only sub-pixel-accurate at the
captured width and leaves a faint ghost elsewhere. remove_watermark_reverse_alpha
now registers the alpha glyph to the actual mark via a TM_CCOEFF_NORMED
scale+position search (_aligned_alpha_map) before inverting the blend, so the
single 2048 capture works at any resolution -- verified clean on the 1773x2364
(3:4) corpus size, the biggest coverage gap (23 files).
reverse_alpha_available is now just 'asset present' (no width band); the registry
still gates removal on detect so a clean corner is never touched. Drops the
_ALPHA_WIDTH_TOLERANCE gate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(doubao): keep native recovery exact -- fixed geometry at captured width
Integer-pixel NCC alignment landed ~1px off at the captured width, degrading the
otherwise-exact native reverse-alpha (synthetic recovery error 0.94 -> 1.39).
remove_watermark_reverse_alpha now uses exact width-relative geometry within
_ALPHA_NATIVE_BAND of the captured width and the NCC search only off it -- best
of both: native back to 0.94, other resolutions still aligned.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(doubao): harden alignment -- try fixed+aligned, keep least residual (56/56)
On a faint/busy-background mark the NCC alignment peak can wander a few px off
the true mark and leave a residual (2/56 real corpus files). Off the captured
width, remove_watermark_reverse_alpha now builds BOTH the fixed-geometry and the
NCC-aligned alpha map, applies each, and keeps whichever leaves the least
residual mark (re-detect confidence on the bare reverse-alpha) -- geometry wins
on faint marks, alignment on clear ones, no magic threshold. Real-file round-trip
now removes 56/56 detected Doubao clean across every corpus resolution (was 54).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* perf(doubao): skip residual inpaint at native width for exact recovery
At the captured width the fixed-geometry reverse-alpha is pixel-exact, so
inpainting over it only replaced exactly-recovered interior pixels with a
cv2 hallucination -- measured worse on a textured background (native error
vs true bg 1.6 reverse-alpha-only vs 2.6 with the old always-on
full-footprint inpaint). Native now returns the bare recovery untouched;
off-native, where NCC alignment is only sub-pixel-approximate, the footprint
inpaint stays to clean the seam. Real round-trip still 56/56 across all
corpus resolutions; negatives 0/60, Gemini unaffected.
Add test_native_returns_exact_reverse_alpha_no_inpaint as the regression
guard. Sync CLAUDE.md + README (the table cell and prose described the
pre-NCC "skipped off native / cv2-LaMa" behavior, now stale). Gitignore the
session scheduled_tasks.lock, and add the text-protection research note.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The text-protection detector scaled every image to a fixed 736 px long side, so
small text on large canvases (e.g. ~16 px on 2048) was downscaled below the
detector and missed -> deformed by the SDXL pass (issue #14). Detect at the
native long side capped at 1536, never upscaled (_detection_input_size, a pure
unit-tested helper). Detection is script-agnostic (DB segments regions, not
characters), so this is language-agnostic: a new benchmark
(scripts/text_detection_benchmark.py) measures recall across Latin/Cyrillic/CJK/
Hangul/Arabic/digits x sizes x canvas -> overall hit-rate 0.91 -> 1.00, worst
cell (2048/16 px) 0.06 -> 1.00. Docs updated.
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(device): support xpu backend
* Fall back to CPU seed generator when device RNG unsupported (xpu)
Some torch-xpu builds have no device-side RNG, so torch.Generator(device="xpu")
raises when --seed is used. _make_seed_generator tries the device generator and
falls back to a backend-agnostic CPU generator. Adds a fallback unit test.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Victor Kuznetsov <kuznetsov.va@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* Add cross-platform CI test matrix, PyPI classifiers
CI: new test.yml runs lint (ubuntu) + a test matrix (ubuntu/macos/windows
x py3.10/3.12, core+dev, GPU tests skip) on push to main and PRs, closing the
gap where only the release publish.yml ran (ubuntu, no tests). Add PyPI
classifiers (OS/Python/topic). README Tests badge, CLAUDE.md CI note.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* Make availability tests reflect installed deps, not assume gpu extra
The new core+dev CI matrix has no diffusers, so the invisible-engine
availability tests (asserting is_available() is True unconditionally) and the
two mocked invisible CLI tests (whose command gates on is_available before the
mock) failed. Assert availability == actual importability of torch+diffusers,
and patch the CLI availability gate so the mocked-engine tests run regardless of
the gpu extra.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
remove_ai_metadata now writes JPEG at quality 95 with 4:4:4 (no chroma
subsampling) instead of the lossy PIL defaults (q75, 4:2:0), and preserves
WebP losslessly instead of silently rewriting it as PNG.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
A bare "unknown" verdict reads as the tool being broken. Print a one-line note
right under the verdict explaining that no locally-readable AI signal was found,
that this is not the same as clean (metadata is often stripped), and that
SynthID-class pixel watermarks have no local detector. The why was previously
only in the caveats section below.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
aigc_label now reads the TC260 label from a raw-JSON `AIGC` PNG tEXt chunk
(as Doubao/ByteDance write it, with no namespaced XMP marker) in addition to
the `<TC260:AIGC>` XMP block, via a shared _parse helper gated on a TC260 field
so a generic AIGC key cannot false-positive. New huggingface_job() reads the
hf-job-id PNG chunk; identify surfaces it as a medium-confidence hf_job signal
(parallel to the visible sparkle, never overriding a hard metadata verdict).
Both wired into has_ai_metadata/get_ai_metadata; the PNG save whitelist already
strips them on removal. Found by auditing 646 corpus originals: 28 AIGC and 3
hf-job files the library previously reported as Unknown.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
SDXL img2img regenerates every pixel, so small text and CJK glyphs deform
at the strengths that defeat SynthID (issue #21). With --protect-text a
CJK-native PP-OCRv3 detector (2.4 MB ONNX, cv2.dnn, no torch, cached on
first use) locates text regions and the pass switches to the SDXL
Differential-Diffusion community pipeline: a per-pixel change map keeps
text regions largely intact while the background is regenerated to strip
the watermark. Gated to the SDXL default model; falls back to plain
img2img with a warning when unavailable.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
HEIF/AVIF store XMP as a meta-box `mime` item whose bytes live in mdat/idat, out
of reach of the top-level uuid/jumb box stripper. An AI-label XMP packet there
(TC260 AIGC, IPTC "Made with AI", IPTC 2025.1) was therefore left in place.
isobmff.blank_ai_xmp_packets locates each XMP packet by its <?xpacket begin ...
end?> delimiters and, if it carries an AI marker (_AI_LABEL_MARKERS), overwrites
it with spaces of the SAME length. Equal length means no box size or iloc offset
shifts -- the coded image stays bit-for-bit intact, the item stays structurally
valid, only the AI label content is destroyed. Plain (non-AI) XMP is left alone,
mirroring the top-level XMP-uuid content match. Wired into remove_ai_metadata's
ISOBMFF branch after strip_c2pa_boxes.
Chosen over exiftool (a non-bundled binary dep) to stay pure-Python and
droplet-compatible; over full iinf/iloc surgery to avoid offset-rewrite
corruption risk. The AI labels we target are all XMP, so this closes the
practical gap. An Exif *item* inside the meta box (rare) still needs iinf/iloc
surgery or exiftool -- documented.
4 new tests (TestMetaBoxXmpBlanking): AI packet blanked (same length, marker
gone, surrounding image bytes intact), plain XMP preserved, no-packet no-op, and
end-to-end remove_ai_metadata on a .heic.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Provenance detection no longer relies on a fixed first-MB read. In a streaming /
non-faststart MP4 the C2PA manifest sits AFTER a multi-megabyte mdat, beyond the
1 MB scan window, so it was missed.
- isobmff.scan_c2pa_region(path): a file-seeking top-level box walker that
returns the payloads of uuid/jumb (provenance) boxes, seeking past mdat by
size without reading it -- works on multi-GB files. Returns b"" for
non-ISOBMFF or on read error. Mirrors the box-size encoding of the existing
in-memory _iter_top_level_boxes (largesize / size==0).
- metadata.scan_head(path, size): the shared input for every C2PA/AIGC/IPTC
byte scan -- first __TEXT __DATA __OBJC others dec hex bytes plus, for ISOBMFF, the late provenance-box
payloads. Behavior-neutral (f.read(size)) for non-ISOBMFF inputs.
- Routed all six metadata scan sites (has_ai_metadata, aigc_label,
iptc_ai_system, synthid_source, exif_generator XMP, get_ai_metadata
soft-binding) and identify's head read through scan_head.
6 new tests: late box found by scan_c2pa_region / scan_head, the fixed window
provably misses it, non-ISOBMFF -> b"", front-placed (faststart) regression.
The remaining gap stays documented: EXIF/XMP stored as items inside the meta
box (AVIF/HEIF stills) still needs meta-box surgery or exiftool.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Surface contradictions between independent provenance signals instead of
collapsing to a single verdict -- a strong tell of spoofed, transplanted, or
laundered metadata. Inspired by arXiv:2603.02378.
Two rules in the new _integrity_clashes helper:
- Conflicting AI-origin attributions: two or more distinct AI vendors named by
independent generator stamps (e.g. a C2PA OpenAI manifest on an image whose
EXIF says Make="Ideogram AI").
- Camera + AI: a camera-capture C2PA device (Pixel/Leica/Sony/Nikon/Truepic)
coexisting with an AI-generation marker -- a genuine capture is not AI.
High-precision by design: only hard generator stamps feed it (C2PA issuer when
the source is AI, SynthID proxy, EXIF/XMP generator, IPTC AISystemUsed, xAI,
AIGC). The fuzzy visible sparkle and the open invisible watermark are excluded
-- the latter can be a by-product of our own SDXL removal pass. Vendor
normalization (_vendor_of over _AI_VENDOR_TOKENS) keeps consistent signals from
clashing (C2PA "Google (Gemini)" + SynthID-Google agree); the C2PA vendor is
read from the issuer attribution, not the resolved platform, so a camera label
like "Google Pixel" cannot mis-normalize to an AI vendor.
Surfaced as ProvenanceReport.integrity_clashes (red in the table view, included
in --json). 19 new tests; all real single-origin fixtures (chatgpt/firefly/
doubao/grok/mj) verified to produce zero clashes (false-positive guard).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two CLI/IO robustness bugs surfaced by issues #17 and #19.
#17 -- non-ASCII image paths (Chinese/Cyrillic/accented) failed on Windows:
cv2.imread/imwrite use the platform ANSI code-page API, so the decode came back
empty with a "can't open/read file" warning. New image_io.imread/imwrite route
through np.fromfile+cv2.imdecode / cv2.imencode+tofile (Unicode-safe, byte-
identical output, cv2.imread None-semantics preserved); all 8 cv2 read/write
call sites now go through it. Behavior-neutral on macOS/Linux (already accept
UTF-8 paths), so the fix is correct-by-construction for the Windows-only bug.
#19 (incidental) -- rich parsed the "[gpu]" in the GPU-extra install hint as a
style tag and dropped it, so the printed command was the un-installable
"pip install 'remove-ai-watermarks'". Escaped as \[gpu] at both call sites.
Tests: test_image_io.py (non-ASCII round-trip, alpha, missing/empty/garbage
semantics); test_cli.py::TestGpuHintMarkup (install hint keeps the extra).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Prevents an unmapped C2PA device whose manifest incidentally contains a mapped
issuer substring (e.g. the "Adobe XMP" toolkit string in a Canon/Sony camera
capture) from being mislabeled as that AI generator ("Adobe Firefly").
_attribute_platform now names a specific AI-generator platform only when the
digital-source-type is trainedAlgorithmicMedia; otherwise it degrades to the
neutral "C2PA signer: X" label. Real Firefly/OpenAI/Google output carries the
AI source-type and is unaffected (verified: chatgpt-1.png->OpenAI,
firefly-1.png->Adobe Firefly still attribute). Closes the only real downside of
leaving Canon/Samsung/Bria device signers unmapped: detection and removal were
already unaffected; now the platform label degrades gracefully too.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
remove_ai_metadata now handles non-ISOBMFF audio/video (which the box walker
can't reach) by shelling out to ffmpeg with a lossless stream copy
(`-map_metadata -1 -map_chapters -1 -c copy`): codec data is untouched, only
container tags/chapters (ID3 / RIFF / Vorbis comments / EBML tags) are dropped.
Requires ffmpeg on PATH; raises a clear RuntimeError if absent or if ffmpeg
can't parse the input (instead of crashing in the image path).
Verified end-to-end: a real ffmpeg-made WAV/MP3 with a "Suno AI" title tag ->
tag gone, audio bytes preserved.
NOT built (evaluated, deliberate): Resemble PerTh audio *detection* --
`get_watermark()` returns a raw bit array with no presence/confidence flag, so
reliably telling watermarked from clean needs Resemble's fixed payload or a
confidence API (neither public; no real sample to calibrate). Same wall as the
SynthID pixel detector. AVIF/HEIF meta-box EXIF/XMP stripping also stays a gap
(needs exiftool, a non-installed binary). Both documented in CLAUDE.md.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Victor Kuznetsov