Version: 1.0
Date: 2026-02-11
Purpose: YouTube Content-ID claim prevention, detection, and dispute protocol
YouTube Content-ID matches audio fingerprints, not copyright ownership.
Scenario:
1. Label X recorded Beethoven's Moonlight Sonata → registered fingerprint in Content-ID
2. We rendered Moonlight Sonata independently → uploaded to YouTube
3. Content-ID flags similarity between two recordings of the same composition
4. The claim is on their recording matching ours — NOT on the composition itself
5. Since our recording is independently produced, the claim is invalid
This is expected behavior, not an error. The dispute process exists for this.
Before uploading audio to YouTube:
If all boxes checked → safe to upload.
1. Check YouTube Studio → Content → Copyright claims
2. If no claim → done
3. If claim appears → proceed to Section 4
1. Review all uploaded videos with parksy-audio BGM
2. Check for new claims that appeared after initial 48hr window
3. Log any claims in the Claim Log below
| Field | What to Check | |
| Claimant | Who filed the claim? (label name) | |
|---|---|---|
| Asset | What recording are they claiming? | |
| Match type | Audio fingerprint match | |
| Our file | Which parksy-audio release was flagged? |
Gather from our Recovery Kit:
1. WHITEPAPER_AUDIO_IP.md — Two-layer model explanation
2. MIDI-SOURCES.md — Our MIDI source (self-composed or PD transcription)
3. LICENSE-INSTRUMENTS.md — Our instrument is properly licensed
4. COMPOSERS_PUBLIC_DOMAIN.md — Composition is public domain
5. Rendering proof — Git commit history showing our independent production
YouTube Studio → Select video → Actions → Dispute
Dispute text template:
This audio was independently produced by parksy-audio using:
- Composition: [title] by [composer] (public domain, died [year])
- MIDI: [self-composed / self-transcribed from IMSLP score]
- Instrument: [SoundFont name] ([license type])
- Rendering: FluidSynth, rendered on [date]
This is an original sound recording. The claimant's recording
is a separate performance of the same public domain composition.
Our recording shares no samples, performance data, or audio
from the claimant's version.
Documentation: https://github.com/dtslib1979/parksy-audio/tree/main/docs/legal
Same evidence, more formal language. YouTube reviews the appeal.
If appeal rejected → consider whether the revenue is worth further action.
Cost-benefit rule: If the video makes less than ₩10,000/month, let it go. Focus on new content.
| # | Date | Video | Claimant | Our File | Status | Resolution | ||||||
1. Avoid the most-recorded pieces on YouTube (Für Elise, Canon in D, Moonlight Sonata) — higher chance of fingerprint collision
2. Use less common movements — Beethoven Sonata No.32 gets fewer claims than Moonlight
3. Add subtle uniqueness — Slightly different tempo, dynamic curve, or articulation reduces fingerprint match probability
4. Upload as unlisted first — Check for claims before making public
5. Off-platform lane — For high-risk pieces, use only in local/off-platform contexts (no Content-ID there)
| Composition | Content-ID Risk | Recommended Lane | ||
| Rarely recorded (obscure Schubert, Scriabin) | Low | Platform (YouTube) | ||
|---|---|---|---|---|
| Frequently recorded (Moonlight, Canon in D) | High | Off-Platform first, test Platform | ||
| Self-composed MIDI with PD harmony | Very Low | Platform | ||
| Complex orchestral (many instruments) | Medium | Test unlisted first |
Built between shifts.