A free, open source, and extensible speech-to-text application that works completely offline.
use anyhow::Result;
use hound::{WavReader, WavSpec, WavWriter};
use log::debug;
use std::path::Path;
/// Read a WAV file and return normalised f32 samples.
pub fn read_wav_samples<P: AsRef<Path>>(file_path: P) -> Result<Vec<f32>> {
let reader = WavReader::open(file_path.as_ref())?;
let samples = reader
.into_samples::<i16>()
.map(|s| s.map(|v| v as f32 / i16::MAX as f32))
.collect::<Result<Vec<f32>, _>>()?;
Ok(samples)
}
/// Verify a WAV file by reading it back and checking the sample count.
pub fn verify_wav_file<P: AsRef<Path>>(file_path: P, expected_samples: usize) -> Result<()> {
let reader = WavReader::open(file_path.as_ref())?;
let actual_samples = reader.len() as usize;
if actual_samples != expected_samples {
anyhow::bail!(
"WAV sample count mismatch: expected {}, got {}",
expected_samples,
actual_samples
);
}
Ok(())
}
/// Save audio samples as a WAV file
pub fn save_wav_file<P: AsRef<Path>>(file_path: P, samples: &[f32]) -> Result<()> {
let spec = WavSpec {
channels: 1,
sample_rate: 16000,
bits_per_sample: 16,
sample_format: hound::SampleFormat::Int,
};
let mut writer = WavWriter::create(file_path.as_ref(), spec)?;
// Convert f32 samples to i16 for WAV
for sample in samples {
let sample_i16 = (sample * i16::MAX as f32) as i16;
writer.write_sample(sample_i16)?;
}
writer.finalize()?;
debug!("Saved WAV file: {:?}", file_path.as_ref());
Ok(())
}