SnapOtter/tests/fixtures/LICENSES.md

# Fixture Provenance & Licensing

All fixtures in `tests/fixtures/` must have verified provenance. Only CC0, CC-BY,
CC-BY-SA, and public-domain assets are permitted in the final registry. Assets
still marked `UNVERIFIED-REVIEW` require human sign-off before the license hard
gate (Phase 2b) can close.

## Verified CC0 Assets (Category A: Regenerated / Category B: Replaced)

These assets are generated by project scripts (`gen-synthetic-content.mjs`) and
carry no third-party copyright. They are deterministic and reproducible.

| Path | Category | Generator | Notes |
|------|----------|-----------|-------|
| `content/qr-code.png` | A | qrcode lib | Encodes `https://snapotter.com`, 400x400 |
| `content/qr-code.svg` | A | qrcode lib | Same data, SVG output |
| `content/qr-code.avif` | A | qrcode + Sharp AVIF | Same data, AVIF format |
| `content/barcode.png` | A | Code 128B encoder + Sharp | Encodes `SNAPOTTER-TEST-123`, 699x152 |
| `content/barcode.avif` | A | Code 128B encoder + Sharp | Same barcode, AVIF format |
| `content/ocr-clean.png` | A | Sharp SVG text | "The quick brown fox 12345", 800x200 |
| `content/ocr-japanese.png` | A | Sharp SVG text | Japanese writing passage, 480x172 |
| `content/alt-2page.pdf` | A | PDF byte generator | Minimal 2-page PDF with text |
| `content/multipage-6.pdf` | A | PDF byte generator | Minimal 6-page PDF with text |
| `content/audio-with-tags.mp3` | A | ffmpeg sine 440Hz | 1.2s MP3, ID3: "Test Song" / "Test Artist" |
| `content/video-with-meta.mp4` | A | ffmpeg color frame | 1s, 64x64, solid orange H.264 |
| `content/media-30s.mp4` | A | ffmpeg testsrc2 + sine | 30s, 640x360, test pattern + 440Hz tone |
| `content/media-30s.wav` | A | ffmpeg sine 440Hz | 30s, 44100Hz, mono PCM |
| `content/animated-simpsons.gif` | B | ffmpeg testsrc | 3s test pattern GIF; **replaced** Homer Simpson clip (Fox/Disney copyright) |
| `content/svg-logo.svg` | B | hand-written SVG | Geometric test fixture; **replaced** ConvertICO brand logo (third-party trademark) |
| `content/multi-face.webp` | B | Sharp oval grid | Face-placeholder grid; **replaced** Shutterstock stock photo #199321457 (watermarked, copyrighted) |

## Synthetic / Project-Generated Assets (pre-existing, not in content/)

The following were generated by project scripts or test infrastructure and carry
no third-party copyright concern:

- Root synthetics: `test-200x150.png`, `test-100x100.jpg`, `test-50x50.webp`,
  `test-100x100.svg`, `test-200x150.heic`, `test-1x1.png`, `test-blank.png`,
  `test-portrait-tall.png`, `test-portrait-extreme.png`, `test-with-exif.jpg`,
  `test-scene.png`, `test-fake-transparency.png`, `test-3page.pdf`
- `media/tiny.*` (29 formats), `documents/tiny.*` (17 formats), `data/tiny.*` (9 formats)
- `hostile/*` (generated via `scripts/generate-hostile-fixtures.mjs`)
- `security/*` (SVG XXE test vectors)
- `formats/sample.*` (format samples, 35 files)

## NEEDS HUMAN LICENSE SIGN-OFF (Category C: UNVERIFIED-REVIEW)

The following `content/` assets have unverifiable provenance. They were added
during the test coverage expansion (commit `babca4cf`) or QA bug fixes (commit
`d8cf979d`) without recorded source URLs. They are real-world photos, audio, or
video that require human review to either:

1. Verify they are CC0/CC-BY and record the source, OR
2. Replace them with properly-licensed alternatives.

**Do NOT delete these.** Depth tests (face detection, colorization, red-eye
removal, transcription, stress testing) depend on real-world content.

| Path | Size | Content Description | Why Review Is Needed |
|------|------|---------------------|---------------------|
| `content/portrait-bw.jpeg` | 2.7 MB | B&W portrait of a woman (3775x5662) | Real photo, no EXIF source. Used for colorization tests. |
| `content/portrait-color.jpg` | 89 KB | Business portrait, man in suit (572x1024) | Likely AI-generated (has characteristic artifacts) but unconfirmed. Used for face/portrait tests. |
| `content/portrait-color-dup.jpg` | 89 KB | Identical to portrait-color.jpg | Deliberate duplicate for dedup testing. Same license concern. |
| `content/portrait-headshot.heic` | 1.5 MB | HEIC portrait headshot | No EXIF source. Used for HEIC format tests with face content. |
| `content/portrait-isolated.png` | 81 KB | Same subject as portrait-color, background removed | Derived from portrait-color.jpg. Same license concern. |
| `content/red-eye.jpg` | 369 KB | Woman with red-eye effect (800x460) | Canon EOS 400D DIGITAL, Adobe Photoshop CS2 (2007). Real photo, no source URL. Used for red-eye removal tool. |
| `content/cross-format-chat.webp` | 182 KB | Mobile chat app screenshot (1440x3120) | Screenshot of "HeartCode" bot. Contains personal/third-party UI. Used in color-palette extraction test. |
| `content/ocr-chat.jpeg` | 1.0 MB | Same chat screenshot as above (JPEG) | Same content as cross-format-chat.webp, different format. Used for OCR tests. |
| `content/ocr-scanned.pdf` | 1.0 MB | Scanned document PDF | Unknown scanned content. Used for OCR-on-PDF tests. |
| `content/stress-large.jpg` | 6.7 MB | Muay Thai boxing match (4000x3000) | Samsung Galaxy S24 Ultra photo (2026-04-11). Likely personal photo but contains identifiable people and event branding. Used for large-image stress tests. |
| `content/watermark.jpg` | 87 KB | Mountain lake landscape with text watermark (720x480) | Real landscape photo with "Text Watermark Sample" overlay. Unknown source. Used for watermark removal tests. |
| `content/motorcycle.heif` | 224 KB | HEIF image | Content not inspectable without HEIF decoder. Used for HEIC/HEIF format processing tests. |
| `content/speech-10s.mp4` | 2.3 MB | 10-second video (640x360) | Likely contains speech for transcription tests. Content unverifiable without playback. |
| `content/speech-10s.wav` | 320 KB | 10-second audio (16kHz mono) | Likely contains speech for transcription tests. Content unverifiable without playback. |

### Recommended Actions for Sign-Off

- **Portraits** (5 files): If AI-generated, confirm the generator and its license terms. If stock photos, identify the source. Consider replacing with CC0 portraits from a known source.
- **Chat screenshots** (2 files): Replace with a project-generated screenshot or a CC0 text image. The chat UI belongs to a third-party app.
- **Red-eye photo**: Source the original from the photographer or replace with a CC0 red-eye portrait.
- **Stress-large photo**: If this is a project contributor's personal photo, have them release it under CC0. Otherwise replace.
- **Watermark landscape**: Source the underlying landscape photo. The watermark overlay was likely added by the project.
- **Speech files** (2 files): Confirm content via playback. If synthetic (ffmpeg-generated), mark CC0. If real speech, identify speaker and get consent or replace.
- **Motorcycle HEIF**: Inspect via HEIF viewer and determine source.
- **OCR-scanned PDF**: Inspect document content and determine source.