aria_kit/audio/
mod.rs

1//! Audio capture: the sources, the recordings they write, and the canonical
2//! 16 kHz mono format shared by everything that reads them.
3//!
4//! [`RecordingReader`] builds on every target. Everything that opens a
5//! microphone, a model, or a file is native.
6
7mod recording;
8mod stop;
9
10#[cfg(not(target_family = "wasm"))]
11mod capture;
12#[cfg(not(target_family = "wasm"))]
13mod resample;
14#[cfg(not(target_family = "wasm"))]
15mod spool;
16#[cfg(all(not(target_family = "wasm"), any(test, feature = "test-support")))]
17pub(crate) mod testing;
18#[cfg(not(target_family = "wasm"))]
19mod wav;
20
21use std::time::Duration;
22
23pub use recording::{CaptureError, RecordedAudio, Recording, RecordingReader, RecordingWriter};
24#[cfg(not(target_family = "wasm"))]
25pub use resample::{ResampleError, Resampler};
26pub use stop::{StopSignal, Stopper};
27
28#[cfg(not(target_family = "wasm"))]
29pub use capture::{
30    AudioCapture, MicrophoneError, RecordingMode, ReplayPace, SilenceDuration,
31    SilenceDurationError, WavReplayCapture, WavReplayError,
32};
33#[cfg(all(feature = "microphone", not(target_family = "wasm")))]
34pub use capture::{
35    Microphone, MicrophoneBuilder, PreparedMicrophone, WakeWordControl, WakeWordEvent,
36    WakeWordMonitor, WakeWordSuspension, check_default_input_device,
37};
38#[cfg(not(target_family = "wasm"))]
39pub use spool::{SpoolError, remove_all_spools};
40#[cfg(all(not(target_family = "wasm"), any(test, feature = "test-support")))]
41pub use testing::{canonical_fixture, fixture_of};
42#[cfg(not(target_family = "wasm"))]
43pub use wav::{WavError, read_wav, write_wav};
44
45pub const CANONICAL_SAMPLE_RATE: u32 = 16_000;
46
47/// The longest audio a spool writes or `from_wav_file` reads.
48#[cfg(not(target_family = "wasm"))]
49pub(crate) const MAX_RECORDING_DURATION: Duration = Duration::from_secs(15 * 60);
50pub const CHANNELS: u16 = 1;
51
52/// What a 16-bit sample divides by to land in -1.0..1.0.
53pub(crate) const PCM_SCALE: f32 = 32_768.0;
54
55/// The nearest 16-bit sample to `sample`, saturating outside -1.0..1.0.
56#[cfg(not(target_family = "wasm"))]
57pub(crate) fn pcm16_from_f32(sample: f32) -> i16 {
58    (sample * PCM_SCALE)
59        .round()
60        .clamp(-PCM_SCALE, PCM_SCALE - 1.0) as i16
61}
62
63pub fn f32_from_pcm16(sample: i16) -> f32 {
64    f32::from(sample) / PCM_SCALE
65}
66
67/// How many canonical samples `duration` spans, rounded down. Saturates at
68/// `u64::MAX`.
69pub const fn canonical_samples(duration: Duration) -> u64 {
70    let samples = samples_at(duration, CANONICAL_SAMPLE_RATE);
71    if samples > u64::MAX as u128 {
72        u64::MAX
73    } else {
74        samples as u64
75    }
76}
77
78/// How long `samples` canonical samples last.
79pub fn canonical_duration(samples: u64) -> Duration {
80    let rate = u64::from(CANONICAL_SAMPLE_RATE);
81    Duration::from_secs(samples / rate)
82        + Duration::from_nanos(samples % rate * 1_000_000_000 / rate)
83}
84
85pub(crate) const fn samples_at(duration: Duration, sample_rate: u32) -> u128 {
86    let rate = sample_rate as u128;
87    duration.as_secs() as u128 * rate + duration.subsec_nanos() as u128 * rate / 1_000_000_000
88}
aria_kit/audio/mod.rs

aria_kit/audio/
mod.rs
