Recording intake — a day of talking becomes a week of landed work

The Power of Recording Intake

The "recording-intake" skill transforms a day's worth of audio recordings into actionable tasks, transcripts, and reports. Every step in this process is designed to ensure that the information extracted is both reliable and useful, turning what could be a pile of forgotten WAV files into a structured plan of action.

The Process

The journey from raw audio to actionable tasks involves several distinct phases, each with its own tools and purposes.

Phase 1 — Transcription and Diarization

The first step involves transcribing and diarizing the recordings using the command:

python diarize_folder.py "<folder>"

This script, located in E:\TE-Code\audio-engine, uses pyannote to automatically detect between one to six speakers in a recording. It's crucial to run python gpu_stack_check.py beforehand to avoid the common pitfall of a torch/torchaudio mismatch, which can lead to zero speaker detection and a useless output. On a 3080 GPU, processing five hours of audio typically takes about an hour.

Phase 2 — Audio Rating

Before delving into any content, each recording is rated for quality with:

python rate_recordings.py "<folder>" --write

This process generates a TRANSCRIPT-QUALITY.md file, evaluating factors like signal level, SNR, and transcript density. The quality grade directly influences the level of correction needed. For example, a D-grade recording results in just a topic list, as opposed to a detailed transcript. This step is crucial for avoiding situations where near-silent audio could produce misleadingly coherent transcripts.

Privacy is Paramount

Handling these recordings requires a strict adherence to privacy protocols. All audio, transcripts, and reports remain in the recordings folder and are never shared in public repositories or prompts. Business facts, when included, must be verifiable from public sources and appropriately cited, distinguishing between "Steve said" and "the footer says".

Making It Actionable

After transcription and rating, the skill ensures that every item has a designated place, either in goal sheets or CRM systems. This meticulous process guarantees that nothing gets lost or misrepresented, allowing me to transform a day's worth of conversations into a week's worth of focused work.

In summary, the recording-intake skill is not just about converting audio to text but about creating a reliable system where every piece of information is traceable and actionable, providing a clear path from discussion to delivery.