feat: capture microphone audio to wav

This commit is contained in:
Evandro committed 2026-06-11 01:02:39 -03:00
1 parent f324566b42
commit 1f6ea8e5db
3 files changed
+226 -10

No files matched your search

+8 -1
View File
@@ -7,12 +7,12 @@ use crate::audio::{RecordingSession, RecordingStatus};
use crate::models::{model_statuses, ModelWithStatus};
use crate::settings::AppSettings;
#[derive(Debug)]
pub struct ManagedAppState {
pub settings: Mutex<AppSettings>,
pub recording: Mutex<RecordingSession>,
pub transcript: Mutex<TranscriptDraft>,
pub model_dir: PathBuf,
pub recordings_dir: PathBuf,
}
#[derive(Debug, Default, Clone, Serialize, Deserialize, PartialEq, Eq)]
@@ -38,6 +38,7 @@ impl Default for ManagedAppState {
recording: Mutex::new(RecordingSession::default()),
transcript: Mutex::new(TranscriptDraft::default()),
model_dir: default_model_dir(),
recordings_dir: default_recordings_dir(),
}
}
}
@@ -76,6 +77,12 @@ fn default_model_dir() -> PathBuf {
.join("models")
}
fn default_recordings_dir() -> PathBuf {
std::env::current_dir()
.unwrap_or_else(|_| std::env::temp_dir())
.join("recordings")
}
#[cfg(test)]
mod tests {
use super::*;
+209 -2
View File
@@ -1,3 +1,13 @@
use std::fs;
use std::io::BufWriter;
use std::path::{Path, PathBuf};
use std::sync::mpsc::{self, Receiver, Sender};
use std::sync::{Arc, Mutex};
use std::thread::{self, JoinHandle};
use cpal::traits::{DeviceTrait, HostTrait, StreamTrait};
use cpal::SampleFormat;
use hound::{SampleFormat as WavSampleFormat, WavSpec, WavWriter};
use serde::{Deserialize, Serialize};
use crate::error::{AppError, AppResult};
@@ -9,11 +19,19 @@ pub enum RecordingStatus {
Recording,
}
#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
#[serde(rename_all = "camelCase")]
type SharedWriter = Arc<Mutex<Option<WavWriter<BufWriter<fs::File>>>>>;
pub struct RecordingSession {
pub status: RecordingStatus,
pub sequence: u64,
active: Option<ActiveRecordingHandle>,
}
struct ActiveRecordingHandle {
path: PathBuf,
stop_tx: Sender<()>,
done_rx: Receiver<AppResult<PathBuf>>,
thread: JoinHandle<()>,
}
impl Default for RecordingSession {
@@ -21,6 +39,7 @@ impl Default for RecordingSession {
Self {
status: RecordingStatus::Idle,
sequence: 0,
active: None,
}
}
}
@@ -44,6 +63,194 @@ impl RecordingSession {
self.status = RecordingStatus::Idle;
Ok(self.sequence)
}
pub fn start_capture(&mut self, recordings_dir: &Path) -> AppResult<PathBuf> {
if self.status == RecordingStatus::Recording {
return Err(AppError::RecordingAlreadyActive);
}
fs::create_dir_all(recordings_dir).map_err(|error| AppError::Io(error.to_string()))?;
self.sequence += 1;
let path = recordings_dir.join(format!("recording-{}.wav", self.sequence));
let thread_path = path.clone();
let (stop_tx, stop_rx) = mpsc::channel();
let (ready_tx, ready_rx) = mpsc::channel();
let (done_tx, done_rx) = mpsc::channel();
let thread = thread::spawn(move || {
let result = run_capture_thread(thread_path, stop_rx, ready_tx);
let _ = done_tx.send(result);
});
match ready_rx.recv() {
Ok(Ok(())) => {
self.status = RecordingStatus::Recording;
self.active = Some(ActiveRecordingHandle {
path: path.clone(),
stop_tx,
done_rx,
thread,
});
Ok(path)
}
Ok(Err(error)) => {
let _ = thread.join();
Err(error)
}
Err(error) => {
let _ = thread.join();
Err(AppError::Audio(format!(
"falha ao iniciar thread de gravação: {error}"
)))
}
}
}
pub fn stop_capture(&mut self) -> AppResult<PathBuf> {
if self.status == RecordingStatus::Idle {
return Err(AppError::NoActiveRecording);
}
let active = self
.active
.take()
.ok_or_else(|| AppError::Audio("gravação ativa sem thread de áudio".to_string()))?;
let expected_path = active.path.clone();
active
.stop_tx
.send(())
.map_err(|error| AppError::Audio(error.to_string()))?;
let result = active
.done_rx
.recv()
.map_err(|error| AppError::Audio(error.to_string()))?;
active
.thread
.join()
.map_err(|_| AppError::Audio("thread de gravação encerrou com panic".to_string()))?;
self.status = RecordingStatus::Idle;
result.map(|path| if path.as_os_str().is_empty() { expected_path } else { path })
}
}
fn run_capture_thread(
path: PathBuf,
stop_rx: Receiver<()>,
ready_tx: Sender<AppResult<()>>,
) -> AppResult<PathBuf> {
let setup = (|| -> AppResult<(cpal::Stream, SharedWriter)> {
let host = cpal::default_host();
let device = host
.default_input_device()
.ok_or_else(|| AppError::Audio("nenhum microfone padrão encontrado".to_string()))?;
let supported_config = device
.default_input_config()
.map_err(|error| AppError::Audio(error.to_string()))?;
let sample_format = supported_config.sample_format();
let config: cpal::StreamConfig = supported_config.into();
let writer = WavWriter::create(
&path,
WavSpec {
channels: config.channels,
sample_rate: config.sample_rate.0,
bits_per_sample: 16,
sample_format: WavSampleFormat::Int,
},
)
.map_err(|error| AppError::Io(error.to_string()))?;
let writer = Arc::new(Mutex::new(Some(writer)));
let writer_for_callback = Arc::clone(&writer);
let error_callback = |error| eprintln!("erro no stream de áudio: {error}");
let stream = match sample_format {
SampleFormat::F32 => device.build_input_stream(
&config,
move |data: &[f32], _| write_f32_samples(data, &writer_for_callback),
error_callback,
None,
),
SampleFormat::I16 => device.build_input_stream(
&config,
move |data: &[i16], _| write_i16_samples(data, &writer_for_callback),
error_callback,
None,
),
SampleFormat::U16 => device.build_input_stream(
&config,
move |data: &[u16], _| write_u16_samples(data, &writer_for_callback),
error_callback,
None,
),
other => {
return Err(AppError::Audio(format!(
"formato de amostra não suportado: {other:?}"
)))
}
}
.map_err(|error| AppError::Audio(error.to_string()))?;
stream.play().map_err(|error| AppError::Audio(error.to_string()))?;
Ok((stream, writer))
})();
let (stream, writer) = match setup {
Ok(active) => active,
Err(error) => {
let _ = ready_tx.send(Err(error.clone()));
return Err(error);
}
};
let _ = ready_tx.send(Ok(()));
let _ = stop_rx.recv();
drop(stream);
let writer = writer
.lock()
.map_err(|_| AppError::Audio("falha ao finalizar arquivo WAV".to_string()))?
.take();
if let Some(writer) = writer {
writer
.finalize()
.map_err(|error| AppError::Io(error.to_string()))?;
}
Ok(path)
}
fn write_f32_samples(samples: &[f32], writer: &SharedWriter) {
if let Ok(mut guard) = writer.lock() {
if let Some(writer) = guard.as_mut() {
for sample in samples {
let value = (sample.clamp(-1.0, 1.0) * i16::MAX as f32) as i16;
let _ = writer.write_sample(value);
}
}
}
}
fn write_i16_samples(samples: &[i16], writer: &SharedWriter) {
if let Ok(mut guard) = writer.lock() {
if let Some(writer) = guard.as_mut() {
for sample in samples {
let _ = writer.write_sample(*sample);
}
}
}
}
fn write_u16_samples(samples: &[u16], writer: &SharedWriter) {
if let Ok(mut guard) = writer.lock() {
if let Some(writer) = guard.as_mut() {
for sample in samples {
let centered = *sample as i32 - i16::MAX as i32 - 1;
let _ = writer.write_sample(centered as i16);
}
}
}
}
#[cfg(test)]
+9 -7
View File
@@ -1,10 +1,11 @@
use std::path::PathBuf;
use tauri::State;
use crate::app_state::{AppSnapshot, ManagedAppState};
use crate::injection::{FallbackInjector, InsertMethod, InsertRequest, InsertResult, TextInjector};
use crate::models::{download_default_model, model_statuses, ModelDownloadResult, ModelWithStatus};
use crate::models::{
default_model_path, download_default_model, model_statuses, ModelDownloadResult, ModelWithStatus,
};
use crate::settings::AppSettings;
use crate::transcription::crispasr::{CrispAsrConfig, CrispAsrTranscriber};
use crate::transcription::{Transcriber, TranscriptionInput, TranscriptionOptions};
@@ -16,11 +17,12 @@ pub(crate) fn get_app_snapshot(state: State<'_, ManagedAppState>) -> Result<AppS
#[tauri::command]
pub(crate) fn start_recording(state: State<'_, ManagedAppState>) -> Result<AppSnapshot, String> {
let recordings_dir = state.recordings_dir.clone();
state
.recording
.lock()
.map_err(|_| "falha ao acessar estado de gravação".to_string())?
.start()
.start_capture(&recordings_dir)
.map_err(String::from)?;
state.snapshot()
@@ -30,11 +32,11 @@ pub(crate) fn start_recording(state: State<'_, ManagedAppState>) -> Result<AppSn
pub(crate) fn stop_recording_and_transcribe(
state: State<'_, ManagedAppState>,
) -> Result<AppSnapshot, String> {
let sequence = state
let audio_path = state
.recording
.lock()
.map_err(|_| "falha ao acessar estado de gravação".to_string())?
.stop()
.stop_capture()
.map_err(String::from)?;
let settings = state
@@ -45,11 +47,11 @@ pub(crate) fn stop_recording_and_transcribe(
let transcriber = CrispAsrTranscriber::new(CrispAsrConfig {
sidecar_name: None,
model_path: None,
model_path: Some(default_model_path(&state.model_dir)),
});
let result = transcriber.transcribe(TranscriptionInput {
audio_path: PathBuf::from(format!("recording-{sequence}.wav")),
audio_path,
options: TranscriptionOptions {
model_id: settings.transcription.model_id,
language: settings.transcription.language,