Audel — Ears for AI Agents 👂

A machine-graded audio feedback loop that coding agents consume to self-correct before claiming done: play → hear → report → fix.

Sibling to AgentVision (eyes) and Verel (brain). AgentVision confirms a video renders; Audel confirms it actually has sound and the narration is right.

AI coding agents are deaf — they generate TTS, soundtracks, web-app sound effects, voice-bot prompts, and audiobooks, and never hear the result. Audel gives them ears: it grades audio/voice/media artifacts and returns a machine verdict (pass/warn/fail) with time-grounded issues, so the agent self-corrects before it claims done.

Status

🚧 Early development. This 0.0.1 release reserves the name on PyPI. The graded API is landing incrementally:

import asyncio
from audel import analyze, load_settings   # coming soon

async def main():
    report = await analyze("dist/intro.mp4", settings=load_settings(audio_backend="local"))
    print(report.verdict, [i.message for i in report.issues])

asyncio.run(main())

Planned surface mirrors AgentVision one-to-one: async analyze / check / watch / render / compute_diff, Brief / IntentClaim, Report / Issue (time-grounded via span) / Conformance / Handoff, an AudioBackend protocol, and CLI / MCP / REST adapters. Light base install (ffmpeg + DSP, no torch); ML behind extras ([render], [asr], [cloud], [clap]).

Install

pip install audel

License

MIT © Amit Patole