7 Commits

Author SHA1 Message Date
Bel LaPointe
d05287fa3d update --stream-* defaults 2023-12-19 10:30:10 -05:00
Bel LaPointe
01be2637ca swap order 2023-12-19 10:26:22 -05:00
Bel LaPointe
226bedb80e add --debug to write a file that can be played with cat /tmp/page.rawf32audio | sox -r 16000 -b 32 -t f32 -e floating-point - -d 2023-12-19 10:25:48 -05:00
Bel LaPointe
6b54e500cd i think my recording has gaps 2023-12-19 09:54:21 -05:00
Bel LaPointe
8603f20a24 break into words but keep more stream head/tail tiebreaking 2023-12-19 09:51:11 -05:00
Bel LaPointe
eee0bf5e65 wip... 2023-12-19 09:30:15 -05:00
Bel LaPointe
15a3f8430a WIP trim the head and tail from text output because low confidence 2023-12-19 09:09:38 -05:00

View File

@@ -5,6 +5,8 @@ use std::time::{Duration, Instant};
use chrono; use chrono;
use clap::Parser; use clap::Parser;
use std::thread; use std::thread;
use std::fs::File;
use std::io::Write;
#[derive(Parser, Debug)] #[derive(Parser, Debug)]
struct Flags { struct Flags {
@@ -14,18 +16,29 @@ struct Flags {
#[arg(long, default_value = "8")] #[arg(long, default_value = "8")]
threads: i32, threads: i32,
#[arg(long, default_value = "1.0")]
stream_retain: f32,
#[arg(long, default_value = "5")] #[arg(long, default_value = "5")]
stream_step: u64, stream_step: u64,
#[arg(long, default_value = "0.6")]
stream_retain: f32,
#[arg(long, default_value = "0.3")]
stream_head: f32,
#[arg(long, default_value = "0.3")]
stream_tail: f32,
wav: Option<String>, wav: Option<String>,
#[arg(long, default_value = "false")]
debug: bool,
} }
fn main() { fn main() {
let flags = Flags::parse(); let flags = Flags::parse();
let w = new_whisper(flags.model, flags.threads, Handler{}).unwrap(); let handler = Handler{
head: flags.stream_head,
tail: flags.stream_tail,
};
let w = new_whisper(flags.model, flags.threads, handler).unwrap();
let stream_retain = (flags.stream_retain * 16_000.0) as usize; let stream_retain = (flags.stream_retain * 16_000.0) as usize;
let stream_step = Duration::new(flags.stream_step, 0); let stream_step = Duration::new(flags.stream_step, 0);
match flags.wav { match flags.wav {
@@ -41,6 +54,10 @@ fn main() {
w.transcribe(&audio_data); w.transcribe(&audio_data);
}, },
None => { None => {
match &flags.debug {
true => { File::create("/tmp/page.rawf32audio").unwrap(); },
false => {},
};
let mut buffer = vec![]; let mut buffer = vec![];
let mut last = Instant::now(); let mut last = Instant::now();
new_listener().listen(move |data: Vec<f32>| { new_listener().listen(move |data: Vec<f32>| {
@@ -51,8 +68,22 @@ fn main() {
Err(msg) => eprintln!("{}", msg), Err(msg) => eprintln!("{}", msg),
}; };
for i in stream_retain..buffer.len() { match &flags.debug {
buffer[i - stream_retain] = buffer[i] true => {
let mut f = File::options().append(true).open("/tmp/page.rawf32audio").unwrap();
let mut wav_data = vec![];
for i in buffer.iter() {
for j in i.to_le_bytes() {
wav_data.push(j);
}
}
f.write_all(wav_data.as_slice()).unwrap();
},
false => {},
};
for i in 0..stream_retain {
buffer[i] = buffer[buffer.len() - stream_retain + i];
} }
buffer.truncate(stream_retain); buffer.truncate(stream_retain);
last = Instant::now(); last = Instant::now();
@@ -139,7 +170,7 @@ impl WhisperEngine {
} }
} }
fn transcribe(&self, data: &Vec<f32>) -> Result<String, WhisperError> { fn transcribe(&self, data: &Vec<f32>) -> Result<Whispered, WhisperError> {
let mut params = FullParams::new(SamplingStrategy::Greedy { best_of: 0 }); let mut params = FullParams::new(SamplingStrategy::Greedy { best_of: 0 });
params.set_no_context(true); params.set_no_context(true);
params.set_n_threads(self.threads); params.set_n_threads(self.threads);
@@ -154,11 +185,13 @@ impl WhisperEngine {
let mut state = self.ctx.create_state()?; let mut state = self.ctx.create_state()?;
state.full(params, &data[..])?; state.full(params, &data[..])?;
let mut result = new_whispered();
let num_segments = state.full_n_segments()?; let num_segments = state.full_n_segments()?;
let mut result = "".to_string();
for i in 0..num_segments { for i in 0..num_segments {
let segment = state.full_get_segment_text(i)?; let data = state.full_get_segment_text(i)?;
result = format!("{} {}", result, segment); let start = state.full_get_segment_t0(i)?;
let stop = state.full_get_segment_t1(i)?;
result.push(data, start, stop);
} }
Ok(result) Ok(result)
@@ -170,12 +203,86 @@ struct AWhisper {
ack: Option<std::sync::mpsc::SyncSender<bool>>, ack: Option<std::sync::mpsc::SyncSender<bool>>,
} }
struct Handler {} #[derive(Clone, Debug)]
struct Whispered {
data: Vec<AWhispered>,
}
#[derive(Clone, Debug)]
struct AWhispered {
data: String,
offset: i64,
length: i64,
}
fn new_whispered() -> Whispered {
Whispered{data: vec![]}
}
fn new_a_whispered(data: String, start: i64, stop: i64) -> AWhispered {
AWhispered{
data: data,
offset: start.clone(),
length: stop - start,
}
}
impl Whispered {
fn to_string(&self) -> String {
let mut result = "".to_string();
for i in 0..self.data.len() {
result = format!("{} {}", result, &self.data[i].data);
}
result
}
fn after(&self, t: &f32) -> Whispered {
let mut result = new_whispered();
self.data
.iter()
.filter(|x| x.offset as f32 >= *t)
.for_each(|x| result.data.push(x.clone()));
result
}
fn before(&self, t: &f32) -> Whispered {
let mut result = new_whispered();
let end = match self.data.iter().map(|x| x.offset + x.length).max() {
Some(x) => x,
None => 1,
};
let t = (end as f32) - *t;
self.data
.iter()
.filter(|x| ((x.offset) as f32) <= t)
.for_each(|x| result.data.push(x.clone()));
result
}
fn push(&mut self, data: String, start: i64, stop: i64) {
let words: Vec<_> = data.split_whitespace().collect();
let per_word = (stop - start) / (words.len() as i64);
for i in 0..words.len() {
let start = (i as i64) * per_word;
let stop = start.clone() + per_word;
self.data.push(new_a_whispered(words[i].to_string(), start, stop));
}
}
}
struct Handler {
head: f32,
tail: f32,
}
impl Handler { impl Handler {
fn on_success(&self, result: String) { fn on_success(&self, result: Whispered) {
eprintln!("{}", chrono::Local::now()); eprintln!("{}: {:?}", chrono::Local::now(), &result);
println!("{}", result); println!("{}", result
.after(&(self.head * 100.0))
.before(&(self.tail * 100.0))
.to_string(),
);
} }
fn on_error(&self, msg: String) { fn on_error(&self, msg: String) {
eprintln!("error: {}", msg); eprintln!("error: {}", msg);