Home / wiltodelta / remove-ai-watermarks
Repository Analysis

wiltodelta/remove-ai-watermarks

Remove visible and invisible AI watermarks and provenance metadata from images and video. Python library and CLI for SynthID, C2PA, EXIF, IPTC, XMP, and common generative-AI marks.

29.2 Moderate AI signal View on GitHub

Analysis Overview

This report presents the forensic synthetic code analysis of wiltodelta/remove-ai-watermarks, a Python project with 5,308 GitHub stars. SynthScan v2.0 examined 60,829 lines of code across 201 source files, recording 1599 pattern matches distributed across 12 syntactic categories. The overall adjusted score of 29.2 places this repository in the Moderate AI signal band.

The scanner applied 160+ deterministic lexical heuristics, multi-line block detectors, abstract syntax tree depth profilers, and a cross-file Jaccard similarity matrix to construct a statistically normalised synthetic code estimate. All matches are individually weighted by severity coefficient and contextual multiplier before summation, and the resulting headline score is temporally discounted to account for the repository's development history relative to the commercial emergence of large language model coding tooling (November 2022 onward).

29.2
Adjusted Score
29.2
Raw Score
100%
Time Factor
2026-08-29
Last Push
5.3K
Stars
Python
Language
60.8K
Lines of Code
201
Files
1.6K
Pattern Hits
2026-08-29
Scan Date
0.07
HC Hit Rate

What These Metrics Mean

Adjusted Score
Primary synthetic code indicator. Raw score normalised per 1,000 lines of code and multiplied by the temporal discount factor. This is the definitive comparative metric — use it to rank repositories by AI authorship density.
Raw Score
The unmodified sum of all severity-weighted, context-multiplied pattern match scores before temporal discounting. Reflects the absolute signal strength independent of when the repository was last active.
Time Factor
The temporal discount multiplier (0–100%) applied to the raw score. Repositories last updated before ChatGPT's launch (Nov 2022) receive a 5% factor. Full signal is only assigned to repositories active in the post-adoption era (Jan 2024+).
Pattern Hits
Total count of individual pattern matches across all files and categories. A high hit count with a low score may indicate a very large codebase with isolated AI snippets; a low count with a high score indicates dense, concentrated AI signatures.
HC Hit Rate
High+Critical pattern hits per file, averaged across the repository. This orthogonal signal catches repositories where a few files are densely packed with high-severity AI tells — a strong indicator even when the normalised score appears moderate due to codebase size.
Lines of Code / Files
Total lines and files analysed. The scanner examines 94 file extensions. These denominators are used to normalise the score, enabling fair comparison between repositories of vastly different sizes.

Score History

This chart maps the temporal evolution of the adjusted synthetic code score across successive scan runs. An upward trajectory indicates ongoing incorporation of AI-generated code or expanding LLM-assisted scaffolding; a stable or declining trajectory may reflect active human refactoring, code removal, or the adoption of stricter authorship policies. The dashed secondary line (right axis) independently tracks total raw pattern hit count, which can diverge from the normalised score when codebase size changes significantly between scans.

Severity Breakdown

Classifies detected patterns by their diagnostic confidence and structural impact. CRITICAL patterns (coefficient 10) represent definitive synthetic signatures — hallucinated imports, explicit LLM attribution metadata — virtually never produced by human authors. HIGH (5) indicates strong structural tells such as cross-file repetition or cross-linguistic idioms. MEDIUM (2) covers recognisable conversational padding and AI-specific vocabulary. LOW (1) captures subtle indicators like tautological comments and generic boilerplate that require density to carry independent signal.

CRITICAL 0HIGH 14MEDIUM 67LOW 1518

Directory Score Breakdown

This horizontal bar chart decomposes the repository's raw synthetic code score by top-level directory, allowing you to pinpoint precisely which modules or components carry the highest AI authorship density. Directories with disproportionately high scores relative to their size warrant targeted manual review: concentrated AI signatures often trace back to mass-generated configuration layers, auto-ported test suites, LLM-scaffolded boilerplate classes, or entire subsystems authored under heavy copilot assistance. Use this view to prioritise your human code-review effort.

Pattern Findings

The scanner identified 1599 distinct pattern matches across 12 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a statistically significant AI authorship indicator. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.

Reading the findings table: The Severity column indicates the diagnostic confidence level (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — comment-context matches receive a ×1.5 weight because LLMs systematically over-annotate. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier as dense clusters constitute far stronger evidence of synthetic authorship than isolated hits.

Hyper-Verbose Identifiers1112 hits · 1070 pts
SeverityFileLineSnippetContext
LOWtests/test_liblib_engine.py46 def test_box_horizontally_centered(self):CODE
LOWtests/test_liblib_engine.py53 def test_box_scales_with_width(self):CODE
LOW⚡tests/test_liblib_engine.py64 def test_strict_only_no_provenance_relaxation(self):CODE
LOW⚡tests/test_liblib_engine.py67 def test_gate_above_clean_arm_max(self):CODE
LOW⚡tests/test_liblib_engine.py71 def test_small_image_size_floor(self):CODE
LOW⚡tests/test_liblib_engine.py85 def test_detects_composed_mark(self):CODE
LOW⚡tests/test_liblib_engine.py91 def test_clean_frame_stays_quiet(self):CODE
LOW⚡tests/test_liblib_engine.py96 def test_mask_covers_logo_and_wordmark(self):CODE
LOW⚡tests/test_liblib_engine.py109 def test_no_mask_on_clean_frame(self):CODE
LOW⚡tests/test_liblib_engine.py114 def test_confident_liblib_detection_suppresses_the_jimeng_pill(self):CODE
LOW⚡tests/test_liblib_engine.py122 def test_force_masks_the_whole_locate_box_on_a_clean_frame(self):CODE
LOW⚡tests/test_gemini_engine.py26 def test_small_image_gets_small_watermark(self):CODE
LOW⚡tests/test_gemini_engine.py29 def test_large_image_gets_large_watermark(self):CODE
LOW⚡tests/test_gemini_engine.py32 def test_boundary_image_stays_small(self):CODE
LOW⚡tests/test_gemini_engine.py40 def test_config_small_returns_correct_values(self):CODE
LOW⚡tests/test_gemini_engine.py46 def test_config_large_returns_correct_values(self):CODE
LOW⚡tests/test_gemini_engine.py52 def test_position_calculation(self):CODE
LOW⚡tests/test_gemini_engine.py65 def test_pure_black_gives_zero_alpha(self):CODE
LOW⚡tests/test_gemini_engine.py71 def test_pure_white_gives_one_alpha(self):CODE
LOW⚡tests/test_gemini_engine.py99 def test_engine_loads_alpha_maps(self):CODE
LOW⚡tests/test_gemini_engine.py105 def test_detect_on_grayscale_does_not_crash(self):CODE
LOW⚡tests/test_gemini_engine.py140 def test_footprint_mask_covers_detected_sparkle(self):CODE
LOW⚡tests/test_gemini_engine.py149 def test_fill_lowers_confidence(self):CODE
LOW⚡tests/test_gemini_engine.py159 def test_footprint_mask_bgra_input(self):CODE
LOW⚡tests/test_gemini_engine.py177 def test_detect_returns_result_object(self, tmp_image_path):CODE
LOW⚡tests/test_gemini_engine.py183 def test_detect_empty_image_returns_no_detection(self):CODE
LOW⚡tests/test_gemini_engine.py189 def test_detect_none_image_returns_no_detection(self):CODE
LOW⚡tests/test_gemini_engine.py193 def test_detect_random_image_low_confidence(self, tmp_image_path):CODE
LOW⚡tests/test_gemini_engine.py239 def test_bright_core_has_high_margin(self):CODE
LOW⚡tests/test_gemini_engine.py245 def test_flat_region_has_low_margin(self):CODE
LOW⚡tests/test_gemini_engine.py255 def test_strong_sparkle_not_demoted(self):CODE
LOW⚡tests/test_gemini_engine.py261 def test_strong_sparkle_on_white_kept(self):CODE
LOWtests/test_gemini_engine.py205 def test_returns_float_in_range_for_real_image(self, tmp_image_path):CODE
LOWtests/test_gemini_engine.py210 def test_returns_none_for_unreadable_file(self, tmp_path):CODE
LOWtests/test_gemini_engine.py273 def test_low_margin_blurred_blob_is_demoted(self):CODE
LOWtests/test_gemini_engine.py295 def test_bright_white_blurred_sparkle_rescued(self):CODE
LOWtests/test_gemini_engine.py319 def test_bright_colored_blob_still_demoted(self):CODE
LOWtests/test_gemini_engine.py379 def test_small_corner_sparkle_is_detected_and_localized(self):CODE
LOWtests/test_gemini_engine.py387 def test_size_weighted_search_alone_traps_on_the_decoy(self):CODE
LOWtests/test_gemini_engine.py410 def test_no_promotion_on_clean_flat_image(self):CODE
LOWtests/test_gemini_engine.py415 def test_low_gradient_decoy_loses_to_high_gradient_corner_sparkle(self, monkeypatch):CODE
LOW⚡tests/test_metadata_internals.py53 def test_supported_formats_include_png(self):CODE
LOW⚡tests/test_metadata_internals.py56 def test_supported_formats_include_jpg(self):CODE
LOW⚡tests/test_metadata_internals.py59 def test_supported_formats_include_heic_avif(self):CODE
LOW⚡tests/test_metadata_internals.py64 def test_supported_formats_exclude_jpeg_xl(self):CODE
LOW⚡tests/test_metadata_internals.py68 def test_ai_metadata_keys_not_empty(self):CODE
LOW⚡tests/test_metadata_internals.py71 def test_ai_keywords_not_empty(self):CODE
LOW⚡tests/test_metadata_internals.py87 def test_extract_metadata_returns_dict(self, tmp_clean_png):CODE
LOW⚡tests/test_metadata_internals.py91 def test_extract_metadata_gets_standard_keys(self, tmp_clean_png):CODE
LOW⚡tests/test_metadata_internals.py95 def test_extract_ai_metadata_from_ai_image(self, tmp_png_with_ai_metadata):CODE
LOW⚡tests/test_metadata_internals.py99 def test_extract_ai_metadata_from_clean_image(self, tmp_clean_png):CODE
LOW⚡tests/test_metadata_internals.py103 def test_has_ai_metadata_detects(self, tmp_png_with_ai_metadata):CODE
LOW⚡tests/test_metadata_internals.py106 def test_has_ai_metadata_clean(self, tmp_clean_png):CODE
LOW⚡tests/test_metadata_internals.py355 def test_current_openai_watermark_action_asserts_synthid(self):CODE
LOW⚡tests/test_metadata_internals.py361 def test_extract_info_adobe_has_no_synthid(self):CODE
LOW⚡tests/test_metadata_internals.py366 def test_extract_chunk_returns_bytes(self):CODE
LOW⚡tests/test_metadata_internals.py399 def test_reader_reports_intact_but_untrusted_credentials(self):CODE
LOW⚡tests/test_metadata_internals.py408 def test_reader_reports_post_signing_container_mutation(self, tampered_chatgpt_png):CODE
LOW⚡tests/test_metadata_internals.py415 def test_fallback_to_png_parser_when_reader_unavailable(self, monkeypatch):CODE
LOW⚡tests/test_metadata_internals.py456 def test_key_not_found_returns_none(self):CODE
1052 more matches not shown…
Decorative Section Separators54 hits · 188 pts
SeverityFileLineSnippetContext
MEDIUM⚡tests/test_gemini_engine.py20# ── WatermarkSize / config helpers ──────────────────────────────────COMMENT
MEDIUM⚡tests/test_gemini_engine.py59# ── Alpha map ───────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_gemini_engine.py89# ── GeminiEngine ────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_gemini_engine.py113# ── Removal: localize -> fill ───────────────────────────────────────COMMENT
MEDIUM⚡tests/test_gemini_engine.py167# ── Detection ───────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_metadata_internals.py47# ── Constants ───────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_metadata_internals.py81# ── Extractor ───────────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_metadata_internals.py545# ── ISOBMFF (AVIF / HEIF / JPEG-XL container stripping) ──────────────COMMENT
MEDIUMtests/test_metadata_internals.py118# ── Cleaner ─────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_metadata_internals.py137# ── C2PA ────────────────────────────────────────────────────────────COMMENT
MEDIUMtests/test_metadata_internals.py435# ── CBOR text extraction (parser internals) ─────────────────────────COMMENT
MEDIUM⚡tests/test_metadata.py56# ── Key detection ───────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_metadata.py84# ── has_ai_metadata / get_ai_metadata ───────────────────────────────COMMENT
MEDIUMtests/test_metadata.py538# ── SynthID-source detection (metadata proxy) ────────────────────────COMMENT
MEDIUMtests/test_metadata.py630# ── remove_ai_metadata ──────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_doubao_engine.py58# ── Detection: alpha-template NCC ───────────────────────────────────COMMENT
MEDIUM⚡tests/test_doubao_engine.py126# ── Alpha asset + footprint mask + localize -> fill removal ─────────COMMENT
MEDIUM⚡tests/test_identify.py840# ── Local diffusion parameters (Stable Diffusion / ComfyUI) ─────────COMMENT
MEDIUM⚡tests/test_identify.py946# ── Visible-sparkle fallback (mocked detector) ──────────────────────COMMENT
MEDIUM⚡tests/test_identify.py1137# ── Caveats and serialization ───────────────────────────────────────COMMENT
MEDIUMtests/test_identify.py341# ── Pure attribution logic (no file IO) ─────────────────────────────COMMENT
MEDIUMtests/test_identify.py620# ── End-to-end verdicts on real fixtures ────────────────────────────COMMENT
MEDIUMtests/test_identify.py874# ── China TC260 AIGC label as a PNG text chunk (Doubao) ─────────────COMMENT
MEDIUMtests/test_identify.py901# ── Hugging Face-hosted job marker (medium confidence) ─────────────COMMENT
MEDIUMtests/test_identify.py1386# ── Open invisible watermark (SD/SDXL/FLUX) integration ─────────────COMMENT
MEDIUMtests/test_identify.py1453# ── Integrity clashes (contradictions between independent signals) ──────COMMENT
MEDIUM⚡tests/test_platform.py31# ── Device detection ────────────────────────────────────────────────COMMENT
MEDIUM⚡tests/test_platform.py367# ── Format utilities ────────────────────────────────────────────────COMMENT
MEDIUMtests/test_platform.py405# ── Availability checks ────────────────────────────────────────────COMMENT
MEDIUMtests/test_platform.py451# ── Platform-specific path handling ─────────────────────────────────COMMENT
MEDIUMscripts/render_visible_examples.py161# ── Video mark examples ──────────────────────────────────────────────────────COMMENT
MEDIUMscripts/fidelity_metrics.py73# ── helpers ──────────────────────────────────────────────────────────COMMENT
MEDIUMscripts/fidelity_metrics.py95# ── text: PaddleOCR defaults ─────────────────────────────────────────COMMENT
MEDIUMscripts/fidelity_metrics.py174# ── face: detection + ArcFace + texture ──────────────────────────────COMMENT
MEDIUMscripts/fidelity_metrics.py246# ── whole image: LPIPS / SSIM / PSNR ─────────────────────────────────COMMENT
MEDIUMscripts/fidelity_metrics.py277# ── reporting ────────────────────────────────────────────────────────COMMENT
MEDIUMscripts/fidelity_metrics.py288# ── CLI ──────────────────────────────────────────────────────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/metadata.py33# ── Known AI metadata keys ──────────────────────────────────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/metadata.py1635# ── Per-file probe memoization ──────────────────────────────────────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/metadata.py1719# ── Shared with the portable metadata record ────────────────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/identify.py1160 # ── C2PA Content Credentials ────────────────────────────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/identify.py1259 # ── SynthID provenance evidence ─────────────────────────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/identify.py1399 # ── Local diffusion parameters (Stable Diffusion / ComfyUI) ──────COMMENT
MEDIUMsrc/remove_ai_watermarks/identify.py1418 # ── xAI / Grok EXIF signature scheme (no C2PA/SynthID/IPTC) ──────COMMENT
MEDIUMsrc/remove_ai_watermarks/identify.py1429 # ── Hugging Face-hosted job marker (hf-job-id PNG text chunk) ─────COMMENT
MEDIUMsrc/remove_ai_watermarks/identify.py1441 # ── Samsung Galaxy AI editing marker (genAIType) ─────────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/identify.py1457 # ── Open invisible watermark (SD / SDXL / FLUX, dwtDct) ──────────COMMENT
MEDIUMsrc/remove_ai_watermarks/identify.py1476 # ── Verdict so far (metadata + embedded watermark) ──────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/humanizer.py83# ── Adaptive polish (target the input's detail level; spare text) ──────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/_text_mark_engine.py289 # ── Templates (delegate to the asset-keyed module cache) ────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/_text_mark_engine.py334 # ── Locate ──────────────────────────────────────────────────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/_text_mark_engine.py523 # ── Mask ────────────────────────────────────────────────────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/_text_mark_engine.py567 # ── Detect ──────────────────────────────────────────────────────────COMMENT
MEDIUMsrc/remove_ai_watermarks/_text_mark_engine.py670 # ── Inpaint footprint (for the inpaint-fallback removal path) ────────COMMENT
Unused Imports162 hits · 156 pts
SeverityFileLineSnippetContext
LOWtests/test_cpu_offload.py9CODE
LOWtests/test_liblib_engine.py9CODE
LOWtests/test_gemini_engine.py3CODE
LOWtests/test_metadata_internals.py4CODE
LOWtests/test_pill_engine.py4CODE
LOWtests/conftest.py3CODE
LOWtests/test_baidu_engine.py11CODE
LOWtests/test_forensic_metadata.py3CODE
LOWtests/test_synthid_pixel_probe.py9CODE
LOWtests/test_registered_mark_calibrate.py3CODE
LOWtests/test_kling_engine.py9CODE
LOWtests/test_packaging.py3CODE
LOWtests/test_cli_robustness.py20CODE
LOWtests/test_jimeng_engine.py8CODE
LOWtests/test_security_clamp.py17CODE
LOWtests/test_microsoft_engine.py9CODE
LOWtests/test_designer_synthid_suppression.py11CODE
LOWtests/test_trustmark_detector.py9CODE
LOWtests/test_video_synthid_sweep.py3CODE
LOWtests/test_synthid_corpus.py3CODE
LOWtests/test_tiling.py9CODE
LOWtests/test_video_temporal.py3CODE
LOWtests/test_qwen_zimage_pipeline.py3CODE
LOWtests/test_fidelity_matching.py12CODE
LOWtests/test_watermark_registry.py3CODE
LOWtests/test_invisible_watermark.py7CODE
LOWtests/test_visible_examples.py12CODE
LOWtests/test_region_eraser.py3CODE
LOWtests/test_video_fidelity_probe.py8CODE
LOWtests/test_metadata.py3CODE
LOWtests/test_video.py3CODE
LOWtests/test_text_mark_faint_mask.py11CODE
LOWtests/test_image_io.py3CODE
LOWtests/test_distribution_boundary.py3CODE
LOWtests/test_script_help.py3CODE
LOWtests/test_text_draft.py3CODE
LOWtests/test_yuanbao_engine.py3CODE
LOWtests/test_invisible_engine.py3CODE
LOWtests/test_doubao_engine.py3CODE
LOWtests/test_infer_text_lines.py1CODE
LOWtests/test_identify.py7CODE
LOWtests/test_qwen_engine.py9CODE
LOWtests/test_contentseal_corpus.py9CODE
LOWtests/test_video_invisible.py3CODE
LOWtests/test_video_invisible.py24CODE
LOWtests/test_optional_deps.py3CODE
LOWtests/test_pixel_evidence.py9CODE
LOWtests/test_selective_text_restoration.py1CODE
LOWtests/test_platform.py7CODE
LOWtests/test_samsung_engine.py9CODE
LOWtests/test_api.py4CODE
LOWtests/test_text_mark_memory.py9CODE
LOWtests/test_cli.py3CODE
LOWtests/test_inpaint_fallback.py9CODE
LOWtests/test_evaluation_data.py3CODE
LOWtests/test_metadata_record.py8CODE
LOWtests/test_runninghub_engine.py11CODE
LOWtests/test_text_restoration.py3CODE
LOWscripts/detection_timing.py45CODE
LOWscripts/video_synthid_sweep.py33CODE
102 more matches not shown…
Excessive Try-Catch Wrapping120 hits · 132 pts
SeverityFileLineSnippetContext
MEDIUMtests/test_metadata_internals.py974def outcome(fn):CODE
LOWtests/test_metadata_internals.py977 except Exception as exc:CODE
LOWtests/test_pill_engine.py20 except Exception:CODE
MEDIUMtests/test_pill_engine.py16def _font_ok() -> bool:CODE
LOWtests/test_image_io.py120 except Exception:CODE
LOW⚡tests/test_image_io.py167 except Exception:CODE
LOWtests/test_video_invisible.py55 except Exception as exc: # pragma: no cover - mutation cleanup pathCODE
MEDIUMtests/test_video_invisible.py51def write_frames() -> None:CODE
LOWscripts/detection_timing.py123 except Exception: # unreadable or an unsupported containerCODE
LOWscripts/detection_timing.py170 except Exception as exc:CODE
LOWscripts/detection_timing.py179 except Exception as exc:CODE
LOWscripts/vendor_mark_calibrate.py101 except Exception:CODE
LOWscripts/vendor_mark_calibrate.py159 except Exception: # noqa: S112 -- one bad file must not kill the sweepCODE
LOWscripts/vendor_mark_calibrate.py329 except Exception: # noqa: S112 -- one bad file must not kill the fitCODE
LOWscripts/vendor_mark_calibrate.py464 except Exception:CODE
LOWscripts/vendor_mark_calibrate.py504 except Exception: # noqa: S112 -- one bad file must not kill the poolCODE
LOWscripts/pill_gate_audit.py93 except Exception as e:CODE
LOWscripts/pill_gate_audit.py182 except Exception: # noqa: S112CODE
LOWscripts/pill_gate_audit.py198 except Exception: # noqa: S112CODE
LOWscripts/pill_gate_audit.py217 except Exception: # noqa: S112CODE
LOWscripts/record_parity_audit.py93 except Exception as exc:CODE
LOWscripts/record_parity_audit.py98 except Exception as exc:CODE
LOWscripts/fill_quality.py219 except Exception: # noqa: S112 -- an unreadable local file is not a candidateCODE
LOWscripts/fill_quality.py271 except Exception as e:CODE
LOWscripts/visible_groundtruth.py68 except Exception:CODE
LOWscripts/synthid_corpus.py97 except Exception: # sips missing/odd output; dims stay unknownCODE
LOWscripts/synthid_corpus.py109 except Exception as exc: # unknown/user formats can raise non-OSError; see CLAUDE.mdCODE
LOWscripts/visible_removal_audit.py109 except Exception as exc:CODE
LOWscripts/detector_response.py196 except Exception: # noqa: S112 -- tolerate a torn last lineCODE
LOWscripts/detector_response.py210 except Exception as e: # a crashed source must not kill the sweepCODE
LOWscripts/visible_positives.py58 except Exception as e:CODE
LOWscripts/visible_positives.py101 except Exception: # noqa: S112 -- tolerate a torn last lineCODE
LOWscripts/visible_positives.py126 except Exception: # noqa: S112CODE
LOWscripts/metadata_removal_audit.py72 except Exception:CODE
LOW⚡scripts/metadata_removal_audit.py100 except Exception as exc:CODE
LOW⚡scripts/metadata_removal_audit.py110 except Exception as exc:CODE
LOW⚡scripts/metadata_removal_audit.py118 except Exception as exc:CODE
LOWscripts/metadata_removal_audit.py140 except Exception: # noqa: S112 -- skip an unreadable identify JSON, not security-relevantCODE
LOWscripts/resource_ceilings.py76except Exception as e:CODE
LOWscripts/corpus_gap_scan.py157 except Exception as exc:CODE
LOWscripts/ladder_headroom.py135 except Exception:CODE
LOWscripts/ladder_headroom.py147 except Exception:CODE
LOWscripts/ladder_headroom.py219 except Exception: # noqa: S112 -- one bad file must not kill the sweepCODE
LOWscripts/vendor_cohort_harvest.py76 except Exception:CODE
LOWscripts/vendor_cohort_harvest.py116 except Exception: # noqa: S112 -- one bad file must not kill the scanCODE
LOWscripts/smoke_matrix.py133 except Exception as e:CODE
LOWscripts/smoke_matrix.py158 except Exception as e:CODE
LOWscripts/smoke_matrix.py397 except Exception:CODE
LOWscripts/sidecar_regression.py133 except Exception as e: # a corrupt sidecar must not kill the sweepCODE
LOWscripts/sidecar_regression.py148 except Exception as e: # record and continue; a crash IS a findingCODE
LOWscripts/sidecar_regression.py187 except Exception: # noqa: S112 -- tolerate a torn last line from an interrupted runCODE
LOW⚡scripts/cjk_tail_probe.py128 except Exception:CODE
LOW⚡scripts/cjk_tail_probe.py136 except Exception:CODE
LOW⚡scripts/cjk_tail_probe.py144 except Exception:CODE
LOWscripts/cjk_tail_probe.py196 except Exception: # noqa: S112 -- one bad file must not kill the sweepCODE
LOWscripts/controlnet_sweep.py279 except Exception as exc:CODE
LOWsrc/remove_ai_watermarks/pixel_evidence.py178 except Exception as exc:CODE
LOWsrc/remove_ai_watermarks/pixel_evidence.py292 except Exception as exc:CODE
LOW⚡src/remove_ai_watermarks/pixel_evidence.py433 except Exception as exc:CODE
LOW⚡src/remove_ai_watermarks/pixel_evidence.py443 except Exception as exc:CODE
60 more matches not shown…
Deep Nesting61 hits · 53 pts
SeverityFileLineSnippetContext
LOWtests/test_pixel_evidence.py167CODE
LOWtests/test_evaluation_data.py14CODE
LOWscripts/vendor_mark_calibrate.py470CODE
LOWscripts/pill_gate_audit.py112CODE
LOWscripts/pill_gate_audit.py167CODE
LOWscripts/selective_text_restoration.py344CODE
LOWscripts/fill_quality.py225CODE
LOWscripts/visible_groundtruth.py80CODE
LOWscripts/synthid_corpus.py77CODE
LOWscripts/visible_removal_audit.py75CODE
LOWscripts/visible_eval.py67CODE
LOWscripts/render_visible_examples.py171CODE
LOWscripts/detector_response.py127CODE
LOWscripts/detector_response.py174CODE
LOWscripts/visible_positives.py79CODE
LOWscripts/real_examples_e2e.py101CODE
LOWscripts/metadata_removal_audit.py180CODE
LOWscripts/detection_timing_report.py147CODE
LOWscripts/fidelity_metrics.py137CODE
LOWscripts/fidelity_metrics.py323CODE
LOWscripts/sidecar_regression.py154CODE
LOWscripts/controlnet_sweep.py225CODE
LOWsrc/remove_ai_watermarks/video_invisible.py318CODE
LOWsrc/remove_ai_watermarks/video_invisible.py353CODE
LOWsrc/remove_ai_watermarks/metadata.py245CODE
LOWsrc/remove_ai_watermarks/metadata.py289CODE
LOWsrc/remove_ai_watermarks/metadata.py462CODE
LOWsrc/remove_ai_watermarks/metadata.py889CODE
LOWsrc/remove_ai_watermarks/metadata.py930CODE
LOWsrc/remove_ai_watermarks/metadata.py1130CODE
LOWsrc/remove_ai_watermarks/metadata.py1440CODE
LOWsrc/remove_ai_watermarks/metadata_record.py222CODE
LOWsrc/remove_ai_watermarks/invisible_engine.py138CODE
LOWsrc/remove_ai_watermarks/video_visible.py1375CODE
LOWsrc/remove_ai_watermarks/cli.py1351CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py199CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py252CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py315CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py402CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py499CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py526CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py590CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py672CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py708CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py769CODE
LOWsrc/remove_ai_watermarks/identify.py202CODE
LOWsrc/remove_ai_watermarks/identify.py1120CODE
LOWsrc/remove_ai_watermarks/identify.py233CODE
LOWsrc/remove_ai_watermarks/video.py573CODE
LOWsrc/remove_ai_watermarks/_internal/ebml.py112CODE
LOWsrc/remove_ai_watermarks/_internal/ebml.py131CODE
LOWsrc/remove_ai_watermarks/_internal/riff.py44CODE
LOWsrc/remove_ai_watermarks/_internal/c2pa.py134CODE
LOWsrc/remove_ai_watermarks/_internal/c2pa.py261CODE
LOWsrc/remove_ai_watermarks/_internal/c2pa.py416CODE
LOWsrc/remove_ai_watermarks/_internal/isobmff.py219CODE
LOWsrc/remove_ai_watermarks/_internal/isobmff.py239CODE
LOWsrc/remove_ai_watermarks/_internal/isobmff.py345CODE
LOWsrc/remove_ai_watermarks/_internal/isobmff.py478CODE
LOWsrc/remove_ai_watermarks/_internal/isobmff.py633CODE
1 more matches not shown…
Synthetic Comment Markers5 hits · 45 pts
SeverityFileLineSnippetContext
HIGH⚡tests/test_metadata_internals.py518 assert "synthid_watermark" not in info # procedural, not AI-generatedCODE
HIGH⚡tests/test_metadata.py1969 # verdict must stay 'unknown', not 'AI-generated'.COMMENT
HIGH⚡tests/test_identify.py1650 # TC260 AIGC label) -- a genuine camera capture is not AI-generated, soCOMMENT
HIGHtests/test_identify.py1777 # A JUMBF-wrapped manifest from a SynthID-pairing signer on AI-generated content:COMMENT
HIGHsrc/remove_ai_watermarks/metadata.py717# without proving the whole image is AI-generated. Scoped to the Samsung editorCOMMENT
Cross-File Repetition9 hits · 45 pts
SeverityFileLineSnippetContext
HIGHtests/test_jimeng_engine.py0a bright blob is not the glyph shape -> low correlation, not detected.STRING
HIGHtests/test_doubao_engine.py0a bright blob is not the glyph shape -> low correlation, not detected.STRING
HIGHtests/test_samsung_engine.py0a bright blob is not the glyph shape -> low correlation, not detected.STRING
HIGHtests/test_jimeng_engine.py0the glyph silhouette correlates with itself, not with a filled block.STRING
HIGHtests/test_doubao_engine.py0the glyph silhouette correlates with itself, not with a filled block.STRING
HIGHtests/test_samsung_engine.py0the glyph silhouette correlates with itself, not with a filled block.STRING
HIGHtests/test_jimeng_engine.py0footprint_mask must not crash on degenerate sizes or non-3-channel inputs.STRING
HIGHtests/test_doubao_engine.py0footprint_mask must not crash on degenerate sizes or non-3-channel inputs.STRING
HIGHtests/test_samsung_engine.py0footprint_mask must not crash on degenerate sizes or non-3-channel inputs.STRING
Over-Commented Block40 hits · 36 pts
SeverityFileLineSnippetContext
LOWpyproject.toml101 # Generation`` (its ``nucleusmoe_image`` pipeline), which only exists inCOMMENT
LOWpyproject.toml141 # TrustMark requires NumPy <2, whose last release has no CPython 3.13+COMMENT
LOWmaintain.sh1#!/usr/bin/env bashCOMMENT
LOWscripts/selective_text_restoration.py1# /// scriptCOMMENT
LOWscripts/infer_text_lines.py1# /// scriptCOMMENT
LOWscripts/render_vendor_silhouettes.py41 # Kling (可灵) stamps a thin light-gray one-line "可灵AI 3.0" bottom-right (anCOMMENT
LOWscripts/fidelity_metrics.py1# /// scriptCOMMENT
LOWscripts/qwen_scrub_prototype.py1# /// scriptCOMMENT
LOWsrc/remove_ai_watermarks/jimeng_engine.py41TOPHAT_DELTA = 12COMMENT
LOWsrc/remove_ai_watermarks/metadata.py101 b"compositeWithTrainedAlgorithmicMedia",COMMENT
LOWsrc/remove_ai_watermarks/metadata.py701 return NoneCOMMENT
LOWsrc/remove_ai_watermarks/metadata.py1521 if source_path.suffix.lower() in _FFMPEG_STRIP_EXTS:COMMENT
LOWsrc/remove_ai_watermarks/metadata_record.py61# opens a blind spot: past the window only ``png_late_metadata``'s ALLOWLIST isCOMMENT
LOWsrc/remove_ai_watermarks/pill_engine.py41# Geometry (fractions of image WIDTH unless noted); top-left corner.COMMENT
LOWsrc/remove_ai_watermarks/watermark_registry.py61# stays strict. Safe: it only escalates where the mark is corroborated.COMMENT
LOWsrc/remove_ai_watermarks/watermark_registry.py81Trust = Literal["strict", "confirmed"]COMMENT
LOWsrc/remove_ai_watermarks/watermark_registry.py221 (the full-frame footprint mask; used by removal)."""COMMENT
LOWsrc/remove_ai_watermarks/watermark_registry.py241 # confidence sparkle path, and the capture-less pill is too weak to attribute.COMMENT
LOWsrc/remove_ai_watermarks/watermark_registry.py341# would report a sparkle while removal declined it, or vice versa, whenever theCOMMENT
LOWsrc/remove_ai_watermarks/watermark_registry.py361# north-west) drops a real sparkle into this band, and the fixed-slot detector demotesCOMMENT
LOWsrc/remove_ai_watermarks/watermark_registry.py381# Precision is flat above 0.42 and collapses below it, and that bottom band alone isCOMMENT
LOWsrc/remove_ai_watermarks/doubao_engine.py41MAX_SATURATION = 55 # max channel spread to count a pixel as "grayish"COMMENT
LOWsrc/remove_ai_watermarks/identify.py81# Stable JSON contract for callers that pass a verdict between services. Bump thisCOMMENT
LOWsrc/remove_ai_watermarks/identify.py481 ai_source_kind: str | None = NoneCOMMENT
LOWsrc/remove_ai_watermarks/identify.py601 PNG has a structured claim_generator; for JPEG/WebP/AVIF/HEIF/JXL theCOMMENT
LOWsrc/remove_ai_watermarks/identify.py721COMMENT
LOWsrc/remove_ai_watermarks/identify.py841# another vendor's engine, e.g. Microsoft+OpenAI or Microsoft+Google; or an editCOMMENT
LOWsrc/remove_ai_watermarks/identify.py881 clashes: list[str] = []COMMENT
LOWsrc/remove_ai_watermarks/identify.py941 """COMMENT
LOWsrc/remove_ai_watermarks/identify.py1161 has_c2pa = bool(info) or c2pa_marker_in(head)COMMENT
LOWsrc/remove_ai_watermarks/identify.py1261 # `synthid_watermark`), then the byte scan for the containers that keep theCOMMENT
LOWsrc/remove_ai_watermarks/_text_mark_engine.py41# genuine mark while killing the icon/thumbnail noise band (<=96 px). Detection isCOMMENT
LOWsrc/remove_ai_watermarks/_text_mark_engine.py61# [0.340,0.400) 91% 73-98% 23COMMENT
LOWsrc/remove_ai_watermarks/_text_mark_engine.py101 max_saturation: floatCOMMENT
LOWsrc/remove_ai_watermarks/_text_mark_engine.py121 # "gray" correlates the silhouette against the raw GRAYSCALE of the locate box:COMMENT
LOW…c/remove_ai_watermarks/_internal/watermark_profiles.py41COMMENT
LOW…c/remove_ai_watermarks/_internal/watermark_profiles.py61# older SDXL pipeline. OpenAI is the easier oracle: the profile already clearedCOMMENT
LOW…c/remove_ai_watermarks/_internal/watermark_profiles.py81COMMENT
LOWsrc/remove_ai_watermarks/_internal/constants.py201# TC260 producer identity -> the mark key whose vendor signs with it now lives on theCOMMENT
LOW…remove_ai_watermarks/_internal/qwen_zimage_pipeline.py61# 20B model runnable on a consumer card at all; at or above it, streaming is pureCOMMENT
Modern Structural Boilerplate23 hits · 23 pts
SeverityFileLineSnippetContext
LOWsrc/remove_ai_watermarks/pixel_evidence.py56logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/video_invisible.py51__all__ = ["build_temporal_reference", "temporal_residual_ratio"]CODE
LOWsrc/remove_ai_watermarks/metadata.py26logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/gemini_engine.py24logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/invisible_watermark.py36logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/metadata_record.py57logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/invisible_engine.py40logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/region_eraser.py46logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/yuanbao_engine.py36logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/text_draft.py30__all__ = [CODE
LOWsrc/remove_ai_watermarks/__init__.py37__all__ = [CODE
LOWsrc/remove_ai_watermarks/runninghub_engine.py58logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/watermark_registry.py45logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/trustmark_detector.py31logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/cli.py129def _setup_logging(verbose: bool) -> None:CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py41__all__ = [CODE
LOWsrc/remove_ai_watermarks/forensic_metadata.py244def _set_jpeg_trailer(result: dict[str, Any], data: bytes, eoi: int) -> None:CODE
LOWsrc/remove_ai_watermarks/identify.py79logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/_text_mark_engine.py31logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/_internal/watermark_remover.py31logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/_internal/c2pa.py27logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/_internal/isobmff.py50logger = logging.getLogger(__name__)CODE
LOWsrc/remove_ai_watermarks/_internal/tiling.py27logger = logging.getLogger(__name__)CODE
AI Slop Vocabulary6 hits · 20 pts
SeverityFileLineSnippetContext
MEDIUM⚡tests/test_metadata_internals.py502 assert "OpenAI" in info["issuer"] # issuer byte-search still robustCODE
MEDIUM⚡tests/test_platform.py348 # The more-robust watermark wins -> safer (higher) strength.COMMENT
MEDIUMscripts/video_synthid_sweep.py181 # scaling-factor gate lives there, and the harness that produces the certifiedCOMMENT
MEDIUMscripts/resource_ceilings.py71 # cv2 figure, and it GROWS with the input, so the harness would partly manufactureCOMMENT
MEDIUMsrc/remove_ai_watermarks/pixel_evidence.py58# Analysis resolution. Every statistic here is scale-robust, and a 2048px cap keepsCOMMENT
MEDIUMsrc/remove_ai_watermarks/identify.py613# on real samples), and more robust than parsing the claim generator (whichCOMMENT
AI Structural Patterns6 hits · 6 pts
SeverityFileLineSnippetContext
LOWsrc/remove_ai_watermarks/video_invisible.py353CODE
LOWsrc/remove_ai_watermarks/invisible_engine.py138CODE
LOWsrc/remove_ai_watermarks/video.py488CODE
LOWsrc/remove_ai_watermarks/video.py573CODE
LOWsrc/remove_ai_watermarks/video.py745CODE
LOWsrc/remove_ai_watermarks/_internal/watermark_remover.py182CODE
Fake / Example Data1 hit · 1 pts
SeverityFileLineSnippetContext
LOW⚡tests/test_metadata.py981 assert xai_signature(_grok_jpeg(tmp_path, artist="John Doe")) is FalseCODE