Function pyannote_rs::segment

source ยท 
pub fn segment<P: AsRef<Path>>(
    samples: &[i16],
    sample_rate: u32,
    model_path: P,
) -> Result<Vec<Segment>>
Examples found in repository?
examples/save_segments.rs (line 38)
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59

fn main() {
    let audio_path = std::env::args().nth(1).expect("Please specify audio file");

    let segmentation_model_path = "segmentation-3.0.onnx";

    let (samples, sample_rate) = pyannote_rs::read_wav(&audio_path).unwrap();

    let segments = pyannote_rs::segment(&samples, sample_rate, segmentation_model_path).unwrap();

    // Create a folder with the base name of the input file
    let file_stem = Path::new(&audio_path)
        .file_stem()
        .unwrap()
        .to_str()
        .unwrap();
    let output_folder = format!("{}_segments", file_stem);
    fs::create_dir_all(&output_folder).unwrap();

    for segment in segments {
        // Compute the embedding result
        // Save the segment to a .wav file
        let segment_file_name = format!(
            "{}/start_{:.2}_end_{:.2}.wav",
            output_folder, segment.start, segment.end
        );
        write_wav(&segment_file_name, &segment.samples, sample_rate).unwrap();
        println!("Created {}", segment_file_name);
    }
}
More examples
Hide additional examples
examples/infinite.rs (line 22)
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53

fn main() {
    let audio_path = std::env::args().nth(1).expect("Please specify audio file");
    let search_threshold = 0.5;

    let embedding_model_path = "wespeaker_en_voxceleb_CAM++.onnx";
    let segmentation_model_path = "segmentation-3.0.onnx";

    let (samples, sample_rate) = pyannote_rs::read_wav(&audio_path).unwrap();
    let mut embedding_extractor = EmbeddingExtractor::new(embedding_model_path).unwrap();
    let mut embedding_manager = EmbeddingManager::new(usize::MAX);

    let segments = pyannote_rs::segment(&samples, sample_rate, segmentation_model_path).unwrap();

    for segment in segments {
        // Compute the embedding result
        let embedding_result: Vec<f32> = match embedding_extractor.compute(&segment.samples) {
            Ok(result) => result.collect(),
            Err(error) => {
                println!(
                    "Error in {:.2}s: {:.2}s: {:?}",
                    segment.start, segment.end, error
                );
                println!(
                    "start = {:.2}, end = {:.2}, speaker = ?",
                    segment.start, segment.end
                );
                continue; // Skip to the next segment
            }
        };

        // Find the speaker
        let speaker = embedding_manager
            .search_speaker(embedding_result.clone(), search_threshold)
            .ok_or_else(|| embedding_manager.search_speaker(embedding_result, 0.0)) // Ensure always to return speaker
            .map(|r| r.to_string())
            .unwrap_or("?".into());

        println!(
            "start = {:.2}, end = {:.2}, speaker = {}",
            segment.start, segment.end, speaker
        );
    }
}
examples/max_speakers.rs (line 23)
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

fn main() {
    let audio_path = std::env::args().nth(1).expect("Please specify audio file");
    let max_speakers = 6;
    let search_threshold = 0.5;

    let embedding_model_path = "wespeaker_en_voxceleb_CAM++.onnx";
    let segmentation_model_path = "segmentation-3.0.onnx";

    let (samples, sample_rate) = pyannote_rs::read_wav(&audio_path).unwrap();
    let mut embedding_extractor = EmbeddingExtractor::new(embedding_model_path).unwrap();
    let mut embedding_manager = EmbeddingManager::new(max_speakers);

    let segments = pyannote_rs::segment(&samples, sample_rate, segmentation_model_path).unwrap();

    for segment in segments {
        // Compute the embedding result
        let embedding_result = match embedding_extractor.compute(&segment.samples) {
            Ok(result) => result.collect(),
            Err(error) => {
                println!(
                    "Error in {:.2}s: {:.2}s: {:?}",
                    segment.start, segment.end, error
                );
                println!(
                    "start = {:.2}, end = {:.2}, speaker = ?",
                    segment.start, segment.end
                );
                continue; // Skip to the next segment
            }
        };

        // Find the speaker
        let speaker = if embedding_manager.get_all_speakers().len() == max_speakers {
            embedding_manager
                .get_best_speaker_match(embedding_result)
                .map(|r| r.to_string())
                .unwrap_or("?".into())
        } else {
            embedding_manager
                .search_speaker(embedding_result, search_threshold)
                .map(|r| r.to_string())
                .unwrap_or("?".into())
        };

        println!(
            "start = {:.2}, end = {:.2}, speaker = {}",
            segment.start, segment.end, speaker
        );
    }
}