EpochCore Quantum Sound engines live · in-browser

The clone can copy the words. Not the harmonics.

In plain English: your voice is enrolled as a harmonic fingerprint — the partial structure no TTS cloner reproduces — and sealed with a keyed, inaudible watermark. Then we attack it 100 ways: your own re-encodes must trace back to you, and every AI clone — ElevenLabs, XTTS, Bark, OpenVoice — gets flagged. Live, in this page.

① SCREEN
OSCL hyper_daemon
6-layer structural screen before a single byte is decoded
② ENROLL
Voiceprint + keyed mark
STFT harmonic fingerprint · QIM cell mark · catalog match
③ GAUNTLET
100 attacks
re-encodes traced · clones flagged · false accusations counted
01 · Structural screen — every asset gated first
Quantum OSCL · hyper_daemon
armed · idle
02 · Enroll your voiceprint
Drop a track
wav · mp3 · m4a — or use the mic / a sample.
Decoded locally; nothing uploads.
Waiting for a voice — the OSCL daemon screens it before anything decodes.
03 · The clone gauntlet
vs ElevenLabs v3Voxtral · MistralCoqui XTTS-v2Suno BarkMyShell OpenVoiceCohere VoiceOpenAI Voice EngineDeepgram Aura
04 · Audio Mark Studio — mark any standalone track
Drop a track to watermark
wav · mp3 · flac · m4a — OSCL-screened, then keyed harmonic-cell marked.
You get the marked .wav + a stealth-verify sidecar; nothing uploads.
The mark rides the harmonic partials — inaudible, keyed, and detectable after re-encodes. Proof copy is 16 kHz mono; keep your master.
To prove it later: ✓ Verify tab → drop the marked file (or ANY re-encode of it) + the sidecar — the keyed mark is re-detected from the seed.