audio_processing

Stateful, provider-neutral PCM processing utilities for Audio Kit.

The package includes resampling, downmixing, rechunking, metering, adaptive gain, ring buffering, timeline synchronization, mixing, and WAV encoding. Transform state is isolated per track so chunk boundaries do not introduce gaps or reset resamplers.

import 'package:audio_processing/audio_processing.dart';

final meter = AudioMeter();
final reading = meter.process(frame);

AudioRing addresses buffered samples by absolute stream position, which is the vocabulary VAD events use, and AdaptiveGain lifts quiet capture to the levels speech models expect without ever attenuating:

final ring = AudioRing();
ring.append(chunk);
final window = ring.samples(from: speechStart, to: ring.writeIndex);

File-backed WAV recording is opt-in because it uses dart:io:

import 'package:audio_processing/audio_processing_io.dart';

See the Audio Kit repository for pipeline examples.