ZeqSpeech

Synthesis, coding, room acoustics, mastering.

• Protocol ID — zeq-speech
• Category — Audio
• Endpoint — POST /api/audio/speech
• Auth — api-key
• Rate limit — 15/min
• Version — 1.287.0
• Precision — ≤0.1% (KO42-enforced)

What it does

Speech processing with HulyaPulse phoneme alignment. Voice activity detection, speaker diarization, formant analysis with R(t) pitch tracking at 1.287 Hz resolution.

Signature

Request

POST /api/audio/speech

Param	Type	Required	Default	Description
audioData	object	✓	—	Speech audio buffer or reference.
task	string		"vad"	'vad', 'diarization', 'formant', 'pitch', 'emotion'.

Response

{ segments, speakers, formants_Hz, pitch_Hz, zeqond }

Runnable example

curl -sS -X POST \
  -H "Authorization: Bearer $ZEQ_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "audioData": {},
  "task": "vad"
}' \
  "https://api.zeq.dev/api/audio/speech"

Integrate

1. Domain solver — compose with KO42 + two additional operators from the matching family for pulse-coherent results.
2. Digital twin — pipe sensor data into this protocol every Zeqond to keep the model phase-locked with the system.
3. Alert threshold — flag results whose error_pct exceeds 0.1% as out-of-spec events for the operations layer.

Seeds

• Near — wrap /api/audio/speech in a language SDK so builders can call it in three lines.
• Medium — publish a reference integration demonstrating ZeqSpeech alongside a real workload, with pulse-aligned metrics.
• Far — propose ZeqSpeech as an open reference standard so other runtimes can implement it verbatim against the Zeq paper.

Papers

• Zeq paper — https://doi.org/10.5281/zenodo.18158152
• Framework paper — https://doi.org/10.5281/zenodo.15825138

Middleware active. Kernel on the 1.287 Hz HulyaPulse. Awaiting next Zeqond.