harald/wyoming-whisper-rs
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

add changes from whisper.cpp update

Browse source
This commit is contained in:
Zero 2023-04-17 17:57:00 -06:00
parent 7c78c128a1
commit 13d44e5881
No known key found for this signature in database
GPG key ID: 3861E636EA1E0E2B
9 changed files with 536 additions and 140 deletions
Download patch file Download diff file Expand all files Collapse all files

24
examples/audio_transcription.rs
View file

@ -9,11 +9,12 @@ use whisper_rs::{FullParams, SamplingStrategy, WhisperContext};
/// Loads a context and model, processes an audio file, and prints the resulting transcript to stdout.
fn main() -> Result<(), &'static str> {
// Load a context and model.
let mut ctx = WhisperContext::new("example/path/to/model/whisper.cpp/models/ggml-base.en.bin")
let ctx = WhisperContext::new("example/path/to/model/whisper.cpp/models/ggml-base.en.bin")
.expect("failed to load model");
// Create a single global key.
ctx.create_key(()).expect("failed to create key");
// Create a params object for running the model.
// Currently, only the Greedy sampling strategy is implemented, with BeamSearch as a WIP.
// The number of past samples to consider defaults to 0.
let mut params = FullParams::new(SamplingStrategy::Greedy { best_of: 0 });
@ -62,18 +63,27 @@ fn main() -> Result<(), &'static str> {
}
// Run the model.
ctx.full(params, &audio[..]).expect("failed to run model");
ctx.full(&(), params, &audio[..])
.expect("failed to run model");
// Create a file to write the transcript to.
let mut file = File::create("transcript.txt").expect("failed to create file");
// Iterate through the segments of the transcript.
let num_segments = ctx.full_n_segments();
let num_segments = ctx
.full_n_segments(&())
.expect("failed to get number of segments");
for i in 0..num_segments {
// Get the transcribed text and timestamps for the current segment.
let segment = ctx.full_get_segment_text(i).expect("failed to get segment");
let start_timestamp = ctx.full_get_segment_t0(i);
let end_timestamp = ctx.full_get_segment_t1(i);
let segment = ctx
.full_get_segment_text(&(), i)
.expect("failed to get segment");
let start_timestamp = ctx
.full_get_segment_t0(&(), i)
.expect("failed to get start timestamp");
let end_timestamp = ctx
.full_get_segment_t1(&(), i)
.expect("failed to get end timestamp");
// Print the segment to stdout.
println!("[{} - {}]: {}", start_timestamp, end_timestamp, segment);

24
examples/basic_use.rs
View file

@ -7,7 +7,10 @@ use whisper_rs::{FullParams, SamplingStrategy, WhisperContext};
// more dependencies than the base library.
pub fn usage() -> Result<(), &'static str> {
// load a context and model
let mut ctx = WhisperContext::new("path/to/model").expect("failed to load model");
let ctx = WhisperContext::new("path/to/model").expect("failed to load model");
// make a sample key
// here, since we only use this model once, we use a unique global key
ctx.create_key(()).expect("failed to create key");
// create a params object
// note that currently the only implemented strategy is Greedy, BeamSearch is a WIP
@ -41,15 +44,24 @@ pub fn usage() -> Result<(), &'static str> {
)?;
// now we can run the model
ctx.full(params, &audio_data[..])
// note the key we use here is the one we created above
ctx.full(&(), params, &audio_data[..])
.expect("failed to run model");
// fetch the results
let num_segments = ctx.full_n_segments();
let num_segments = ctx
.full_n_segments(&())
.expect("failed to get number of segments");
for i in 0..num_segments {
let segment = ctx.full_get_segment_text(i).expect("failed to get segment");
let start_timestamp = ctx.full_get_segment_t0(i);
let end_timestamp = ctx.full_get_segment_t1(i);
let segment = ctx
.full_get_segment_text(&(), i)
.expect("failed to get segment");
let start_timestamp = ctx
.full_get_segment_t0(&(), i)
.expect("failed to get segment start timestamp");
let end_timestamp = ctx
.full_get_segment_t1(&(), i)
.expect("failed to get segment end timestamp");
println!("[{} - {}]: {}", start_timestamp, end_timestamp, segment);
}

13
examples/full_usage/src/main.rs
View file

@ -45,18 +45,19 @@ fn main() {
let original_samples = parse_wav_file(audio_path);
let samples = whisper_rs::convert_integer_to_float_audio(&original_samples);
let mut ctx =
let ctx =
WhisperContext::new(&whisper_path.to_string_lossy()).expect("failed to open model");
ctx.create_key(()).expect("failed to create key");
let params = FullParams::new(SamplingStrategy::default());
ctx.full(params, &samples)
ctx.full(&(), params, &samples)
.expect("failed to convert samples");
let num_segments = ctx.full_n_segments();
let num_segments = ctx.full_n_segments(&()).expect("failed to get number of segments");
for i in 0..num_segments {
let segment = ctx.full_get_segment_text(i).expect("failed to get segment");
let start_timestamp = ctx.full_get_segment_t0(i);
let end_timestamp = ctx.full_get_segment_t1(i);
let segment = ctx.full_get_segment_text(&(), i).expect("failed to get segment");
let start_timestamp = ctx.full_get_segment_t0(&(), i).expect("failed to get start timestamp");
let end_timestamp = ctx.full_get_segment_t1(&(), i).expect("failed to get end timestamp");
println!("[{} - {}]: {}", start_timestamp, end_timestamp, segment);
}
}