whisperd 0.1.12

HTTP server for the OpenAI Whisper speech-to-text model.
Documentation 
1
2
3
4
5
6
7
8
9
10
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
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257

#![warn(clippy::all, clippy::pedantic, clippy::nursery)]

use clap::Parser;
use hyper::service::{make_service_fn, service_fn};
use hyper::{header::CONTENT_TYPE, Body, Request, Response, Server, StatusCode};
use multer::Multipart;
use serde::Serialize;
use serde_json::to_string;
use std::fs;
use std::path::Path;
use std::sync::{Arc, Mutex};
use std::{convert::Infallible, net::SocketAddr};
use whisperd::{Language, Model, Size, Whisper};

use crate::utils::write_to;

mod utils;

#[derive(Serialize)]
struct TranscriptionResponse {
    text: String,
}

#[derive(Parser)]
struct Opts {
    #[clap(subcommand)]
    subcmd: SubCommand,
}

#[derive(Parser)]
enum SubCommand {
    #[command(about = "Start the transcription server.")]
    Serve {
        /// Port to listen on
        #[clap(short, long, default_value = "8000")]
        port: u16,
        /// Path to the Whisper model
        #[clap(short, long)]
        model_path: String,
    },
    #[command(about = "Transcribe a given audio file.")]
    Transcribe(TranscribeArgs),
}

#[derive(Parser)]
struct TranscribeArgs {
    /// Name of the Whisper model to use
    #[clap(short, long, default_value = "medium")]
    model: Size,

    /// Language spoken in the audio. Attempts to auto-detect by default.
    #[clap(short, long)]
    lang: Option<Language>,

    /// Path to the audio file to transcribe
    #[clap(name = "AUDIO")]
    audio: String,

    /// Toggle translation
    #[clap(short, long, default_value = "false")]
    translate: bool,

    /// Generate timestamps for each word
    #[clap(short, long, default_value = "false")]
    karaoke: bool,

    /// Write transcription results to .txt, .vtt, and .srt files.
    #[clap(short, long, default_value = "false")]
    write: bool,
}

#[tokio::main]
async fn main() {
    let opts = Opts::parse();
    match opts.subcmd {
        SubCommand::Serve { port, model_path } => {
            let model_path = Path::new(&model_path);
            start_server(port, &model_path).await;
        }
        SubCommand::Transcribe(args) => transcribe_audio(args).await,
    }
}

async fn start_server(port: u16, model_path: &Path) {
    // load model
    let whisper = Arc::new(Mutex::new(
        Whisper::from_model_path(model_path, Some(Language::Auto)).await,
    ));

    let make_svc = make_service_fn(move |_conn| {
        let whisper_clone = whisper.clone();
        async move {
            Ok::<_, Infallible>(service_fn(move |req| {
                handle_transcription(req, whisper_clone.clone())
            }))
        }
    });

    // start listening
    let addr = SocketAddr::from(([127, 0, 0, 1], port));
    let server = Server::bind(&addr).serve(make_svc);

    println!("🏃‍♀️ Server running at: {}", addr);
    if let Err(e) = server.await {
        eprintln!("server error: {}", e);
    }
}

// A handler for incoming requests.
async fn handle_transcription(
    req: Request<Body>,
    whisper: Arc<Mutex<Whisper>>,
) -> Result<Response<Body>, Infallible> {
    // Check if the request is a preflight request (OPTIONS method)
    if req.method() == hyper::Method::OPTIONS && req.uri().path() == "/v1/audio/transcriptions" {
        let res = Response::builder()
            .status(StatusCode::OK)
            .header("Access-Control-Allow-Origin", "*")
            .header("Access-Control-Allow-Methods", "POST, OPTIONS")
            .header("Access-Control-Allow-Headers", "Content-Type")
            .body(Body::empty())
            .unwrap();
        return Ok(res);
    }
    // Extract the `multipart/form-data` boundary from the headers.
    let boundary = req
        .headers()
        .get(CONTENT_TYPE)
        .and_then(|ct| ct.to_str().ok())
        .and_then(|ct| multer::parse_boundary(ct).ok());

    // Send `BAD_REQUEST` status if the content-type is not multipart/form-data.
    if boundary.is_none() {
        return Ok(Response::builder()
            .status(StatusCode::BAD_REQUEST)
            .header("Access-Control-Allow-Origin", "*") // Add this for CORS
            .body(Body::from("BAD REQUEST"))
            .unwrap());
    }

    // Process the multipart e.g. you can store them in files.
    let transcription_request = process_multipart(req.into_body(), boundary.unwrap()).await;

    if let Err(err) = transcription_request {
        return Ok(Response::builder()
            .status(StatusCode::INTERNAL_SERVER_ERROR)
            .header("Access-Control-Allow-Origin", "*") // Add this for CORS
            .body(Body::from(format!("INTERNAL SERVER ERROR: {}", err)))
            .unwrap());
    }

    if let Ok(trans_req) = transcription_request {
        let audio = Path::new(trans_req.as_str());
        let transcript = {
            let mut whisper_guard = whisper.lock().unwrap();
            whisper_guard.transcribe(audio, false, false).unwrap()
        };
        println!("time: {:?}", transcript.processing_time);

        let transcript_text = transcript.as_text();
        let response: TranscriptionResponse = TranscriptionResponse {
            text: transcript_text,
        };

        let json_response = to_string(&response).expect("Failed to serialize to JSON");

        let response = Response::builder()
            .header("Access-Control-Allow-Origin", "*") // Add this for CORS
            .body(Body::from(json_response))
            .unwrap();

        return Ok(response);
    }

    Ok(Response::new(Body::from("Success")))
}

// Process the request body as multipart/form-data.
async fn process_multipart(body: Body, boundary: String) -> multer::Result<String> {
    // Create a Multipart instance from the request body.
    let mut multipart = Multipart::new(body, boundary);
    let mut file_path = String::new();

    // Iterate over the fields, `next_field` method will return the next field if
    // available.
    while let Some(mut field) = multipart.next_field().await? {
        if field.name() == Some("file") {
            // Get the field name.
            let name = field.name();

            // Get the field's filename if provided in "Content-Disposition" header.
            let file_name = field.file_name();

            // Get the "Content-Type" header as `mime::Mime` type.
            let content_type = field.content_type();

            println!(
                "Name: {:?}, FileName: {:?}, Content-Type: {:?}",
                name, file_name, content_type
            );
            // Process the field data chunks e.g. store them in a file.
            let mut bytes_len = 0;
            let mut audio_data = Vec::new();
            while let Some(field_chunk) = field.chunk().await? {
                audio_data.extend_from_slice(&field_chunk);
                bytes_len += field_chunk.len();
            }
            println!("Bytes Length: {:?}", bytes_len);
            let file_name_str: &str = field.file_name().as_ref().unwrap_or(&"audio.wav");
            file_path = format!("/tmp/{}", file_name_str); // Adjust as necessary
            fs::write(&file_path, audio_data).expect("Failed to write to file");
            println!("Write the file to {}", file_path);
        }
    }

    Ok(file_path)
}

async fn transcribe_audio(mut args: TranscribeArgs) {
    let audio = Path::new(&args.audio);
    let file_name = audio.file_name().unwrap().to_str().unwrap();

    assert!(audio.exists(), "The provided audio file does not exist.");

    if args.model.is_english_only() && (args.lang == Some(Language::Auto) || args.lang.is_none()) {
        args.lang = Some(Language::English);
    }

    assert!(
        !args.model.is_english_only() || args.lang == Some(Language::English),
        "The selected model only supports English."
    );

    let mut whisper = Whisper::new(Model::new(args.model), args.lang).await;
    let transcript = whisper
        .transcribe(audio, args.translate, args.karaoke)
        .unwrap();
    println!("time: {:?}", transcript.processing_time);

    if args.write {
        write_to(
            audio.with_file_name(format!("{file_name}.txt")),
            &transcript.as_text(),
        );
        write_to(
            audio.with_file_name(format!("{file_name}.vtt")),
            &transcript.as_vtt(),
        );
        write_to(
            audio.with_file_name(format!("{file_name}.srt")),
            &transcript.as_srt(),
        );
    } else {
        println!("");
        println!("🔊 {}", transcript.as_text());
    }
}