Convert Discord voice PCM into VAD-filtered, Whisper-compatible audio.
npx skills add https://github.com/antonyfmunoz/OS --skill voice-pipeline
Please help me install this Agent Skill. Skill: voice_pipeline Source: https://github.com/antonyfmunoz/OS/tree/main/skills/tools/voice_pipeline Command: npx skills add https://github.com/antonyfmunoz/OS --skill voice-pipeline
This Skill solves unreliable speech detection and noisy audio preprocessing by transforming Discord voice PCM into clean, VAD-filtered, Whisper-ready audio using a two-tier approach with deterministic and neural fallbacks.
int16 PCM bytes → float32 normalized [-1, 1] and to maintain exact frame slicing requirements for VAD.Use Cases: streaming Discord voice capture, tuning real-time transcription quality, improving robustness in noisy environments or music-heavy channels, and building or modifying audio buffers before downstream STT.
Use the voice_pipeline Skill to implement Discord audio capture preprocessing, apply two-tier VAD with music filtering, and output WAV/segments suitable for your Whisper STT stage.
High-intent search queries and answers about installing and using this skill.
Voice activity detection filters raw Discord PCM by using Silero VAD for neural confidence scoring and webrtcvad as a deterministic exact-frame fallback. This dual approach reliably isolates speech segments while discarding silence and background noise.
Silero VAD serves as the primary neural confidence-based detector, while webrtcvad acts as a deterministic fallback when the neural path fails or is unavailable. This two-tier approach ensures continuous speech detection even in noisy environments or music-heavy channels.
You can convert int16 PCM bytes to normalized float32 audio using numpy array conversion, scaling values to the [-1, 1] range. This normalization step maintains exact frame slicing requirements required by both Silero VAD and webrtcvad processors.
You can segment individual user utterances from Discord voice capture by implementing per-user audio buffering combined with silence-threshold flushing. VAD filtering isolates speech, and when silence is detected, the buffer flushes the segmented utterance for transcription.
The best way to handle unreliable speech detection in noisy Discord channels is a two-tier VAD approach using Silero VAD for neural inference and webrtcvad for deterministic fallback, combined with librosa music gating to suppress background audio interference.
KAIROS7-team3Integrate OpenAI Whisper STT for voice recognition and ROS2 command processing.
framerslabTranscribe live speech in real time using the OpenAI Whisper API.
wentbackwardClone voices for TTS and transcribe speech via OpenAI-compatible endpoints.
open-velaDetect and quantify clipping, silence insertion, pops, and noise floor in PCM audio.