found the type needed to pass closures with local variables

.gitignore

@@ -2,6 +2,8 @@
 /whisper-cpp-2023/rust.d/target
 /rust-whisper.d/target
 /rust-whisper.d/models
+/target/
+/models/
 snowboy-2022/snowboy
 **/*.git.d
 **/*.wav

.gitmodules

@@ -1,3 +1,3 @@
 [submodule "rust-whisper.d/gitea-whisper-rs"]
-	path = rust-whisper.d/gitea-whisper-rs
+	path = gitea-whisper-rs
 	url = https://gitea.inhome.blapointe.com/bel/whisper-rs.git

Cargo.lock

@@ -63,6 +63,54 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "anstream"
+version = "0.6.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2ab91ebe16eb252986481c5b62f6098f3b698a45e34b5b98200cf20dd2484a44"
+dependencies = [
+ "anstyle",
+ "anstyle-parse",
+ "anstyle-query",
+ "anstyle-wincon",
+ "colorchoice",
+ "utf8parse",
+]
+
+[[package]]
+name = "anstyle"
+version = "1.0.4"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7079075b41f533b8c61d2a4d073c4676e1f8b249ff94a393b0595db304e0dd87"
+
+[[package]]
+name = "anstyle-parse"
+version = "0.2.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "317b9a89c1868f5ea6ff1d9539a69f45dffc21ce321ac1fd1160dfa48c8e2140"
+dependencies = [
+ "utf8parse",
+]
+
+[[package]]
+name = "anstyle-query"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5ca11d4be1bab0c8bc8734a9aa7bf4ee8316d462a08c6ac5052f888fef5b494b"
+dependencies = [
+ "windows-sys",
+]
+
+[[package]]
+name = "anstyle-wincon"
+version = "3.0.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "f0699d10d2f4d628a98ee7b57b289abbc98ff3bad977cb3152709d4bf2330628"
+dependencies = [
+ "anstyle",
+ "windows-sys",
+]
+
 [[package]]
 name = "autocfg"
 version = "1.1.0"
@@ -213,6 +261,46 @@ dependencies = [
  "libloading",
 ]
 
+[[package]]
+name = "clap"
+version = "4.4.10"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "41fffed7514f420abec6d183b1d3acfd9099c79c3a10a06ade4f8203f1411272"
+dependencies = [
+ "clap_builder",
+ "clap_derive",
+]
+
+[[package]]
+name = "clap_builder"
+version = "4.4.9"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "63361bae7eef3771745f02d8d892bec2fee5f6e34af316ba556e7f97a7069ff1"
+dependencies = [
+ "anstream",
+ "anstyle",
+ "clap_lex",
+ "strsim",
+]
+
+[[package]]
+name = "clap_derive"
+version = "4.4.7"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "cf9804afaaf59a91e75b022a30fb7229a7901f60c755489cc61c9b423b836442"
+dependencies = [
+ "heck",
+ "proc-macro2",
+ "quote",
+ "syn 2.0.39",
+]
+
+[[package]]
+name = "clap_lex"
+version = "0.6.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "702fc72eb24e5a1e48ce58027a675bc24edd52096d5397d4aea7c6dd9eca0bd1"
+
 [[package]]
 name = "cmake"
 version = "0.1.50"
@@ -222,6 +310,12 @@ dependencies = [
  "cc",
 ]
 
+[[package]]
+name = "colorchoice"
+version = "1.0.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "acbf1af155f9b9ef647e42cdc158db4b64a1b61f743629225fde6f3e0be2a7c7"
+
 [[package]]
 name = "combine"
 version = "4.6.6"
@@ -335,6 +429,12 @@ version = "0.14.3"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "290f1a1d9242c78d09ce40a5e87e7554ee637af1351968159f4952f028f75604"
 
+[[package]]
+name = "heck"
+version = "0.4.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "95505c38b4572b2d910cecb0281560f54b440a19336cbbcb27bf6ce6adc6f5a8"
+
 [[package]]
 name = "home"
 version = "0.5.5"
@@ -773,6 +873,7 @@ version = "0.1.0"
 dependencies = [
  "byteorder",
  "chrono",
+ "clap",
  "cpal",
  "signal-hook",
  "tokio",
@@ -851,6 +952,12 @@ version = "1.11.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "4dccd0940a2dcdf68d092b8cbab7dc0ad8fa938bf95787e1b916b0e3d0e8e970"
 
+[[package]]
+name = "strsim"
+version = "0.10.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "73473c0e59e6d5812c5dfe2a064a6444949f089e20eec9a2e5506596494e4623"
+
 [[package]]
 name = "syn"
 version = "1.0.109"
@@ -926,6 +1033,12 @@ version = "1.0.12"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3354b9ac3fae1ff6755cb6db53683adb661634f67557942dea4facebec0fee4b"
 
+[[package]]
+name = "utf8parse"
+version = "0.2.1"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "711b9620af191e0cdc7468a8d14e709c3dcdb115b36f838e601583af800a370a"
+
 [[package]]
 name = "walkdir"
 version = "2.4.0"

Cargo.toml

@@ -13,3 +13,4 @@ cpal = "0.15.2"
 signal-hook = "0.3.17"
 byteorder = "1.5.0"
 chrono = "0.4.31"
+clap = { version = "4.4.10", features = ["derive"] }

README.md

@@ -1,29 +0,0 @@
-# stt
-
-## listen on linux
-
-https://wiki.archlinux.org/title/PulseAudio/Examples
-
-```
-10. ALSA monitor source
-
-To be able to record from a monitor source (a.k.a. "What-U-Hear", "Stereo Mix"), use pactl list to find out the name of the source in PulseAudio (e.g. alsa_output.pci-0000_00_1b.0.analog-stereo.monitor). Then add lines like the following to /etc/asound.conf or ~/.asoundrc:
-
-pcm.pulse_monitor {
-  type pulse
-  device alsa_output.pci-0000_00_1b.0.analog-stereo.monitor
-}
-
-ctl.pulse_monitor {
-  type pulse
-  device alsa_output.pci-0000_00_1b.0.analog-stereo.monitor
-}
-
-Now you can select pulse_monitor as a recording source.
-
-Alternatively, you can use pavucontrol to do this: make sure you have set up the display to "All input devices", then select "Monitor of [your sound card]" as the recording source.
-```
-
-```bash
-$ pactl list | grep -A 50 RUNNING | grep -E 'RUNNING|Name:|Monitor Source:' | grep Monitor.Source | head -n 1 | awk '{print $NF}'
-```

models/download_models.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+
+src="https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml"
+cd "$(dirname "$(realpath "$BASH_SOURCE")")"
+
+# Whisper models
+for model in "tiny.en" "base.en" "small.en" "medium.en" "large-v2"; do
+    	test -f ./ggml-$model.bin || wget --quiet --show-progress -O ./ggml-$model.bin "$src-$model.bin"
+done
+
+test -f ./ggml-distil-medium.en.bin || wget https://huggingface.co/distil-whisper/distil-medium.en/resolve/main/ggml-medium-32-2.en.bin?download=true -O ./ggml-distil-medium.en.bin
+
+test -f ./ggml-distil-large-v2.bin || wget https://huggingface.co/distil-whisper/distil-large-v2/resolve/main/ggml-large-32-2.en.bin?download=true -O ./ggml-distil-large-v2.bin

models/testme/Cargo.lock

@@ -0,0 +1,7 @@
+# This file is automatically @generated by Cargo.
+# It is not intended for manual editing.
+version = 3
+
+[[package]]
+name = "testme"
+version = "0.1.0"

models/testme/Cargo.toml

@@ -0,0 +1,8 @@
+[package]
+name = "testme"
+version = "0.1.0"
+edition = "2021"
+
+# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
+
+[dependencies]

models/testme/src/main.rs

@@ -0,0 +1,4 @@
+fn main() {
+   let bytes = include_bytes!("./test.txt");
+   println!("{}", String::from_utf8_lossy(bytes));
+}

models/testme/src/test.txt

@@ -0,0 +1 @@
+../../ggml-tiny.en.bin

rust-whisper.d/models/download_models.sh

@@ -1,9 +0,0 @@
-#!/bin/bash
-
-src="https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml"
-cd "$(dirname "$(realpath "$BASH_SOURCE")")"
-
-# Whisper models
-for model in "tiny.en" "base.en" "small.en" "medium.en"; do
-    	test -f ./ggml-$model.bin || wget --quiet --show-progress -O ./ggml-$model.bin "$src-$model.bin"
-done

rust-whisper.d/src/main.rs

@@ -1,143 +0,0 @@
-use whisper_rs::{WhisperContext, FullParams, SamplingStrategy, WhisperError};
-use cpal::traits::{HostTrait, DeviceTrait, StreamTrait};
-use signal_hook::{iterator::Signals, consts::signal::SIGINT};
-use std::time::{Duration, Instant};
-use chrono;
-
-fn main() {
-   let w = new_whisper(
-      std::env::var("MODEL")
-         .unwrap_or(String::from("../models/ggml-tiny.en.bin")),
-      std::env::var("P")
-         .unwrap_or(String::from("8"))
-         .parse::<i32>().expect("$P must be a number"),
-   ).unwrap();
-   let stream_churn = std::env::var("STREAM_CHURN")
-      .unwrap_or(String::from("0.8"))
-      .parse::<f32>().expect("$STREAM_CHURN must be a number");
-   let stream_step = Duration::new(
-      std::env::var("STREAM_STEP")
-         .unwrap_or(String::from("5"))
-         .parse::<u64>().expect("$STREAM_STEP must be a number"),
-      0,
-   );
-
-   match std::env::var("WAV") {
-      Ok(wav) => {
-         let (header, data) = wav::read(
-            &mut std::fs::File::open(wav).expect("failed to open $WAV"),
-         ).expect("failed to decode $WAV");
-         assert!(header.channel_count == 1);
-         assert!(header.sampling_rate == 16000);
-         let data16 = data.as_sixteen().expect("wav is not 32bit floats");
-         let audio_data = &whisper_rs::convert_integer_to_float_audio(&data16);
-
-         let result = w.transcribe(&audio_data).unwrap();
-         println!("{}", result);
-      },
-      Err(_) => {
-         let host = cpal::default_host();
-         let device = host.default_input_device().unwrap();
-         let cfg = device.supported_input_configs()
-            .unwrap()
-            .filter(|x| x.sample_format() == cpal::SampleFormat::F32)
-            .nth(0)
-            .unwrap()
-            .with_max_sample_rate();
-
-         let channels = cfg.channels();
-         let downsample_ratio = cfg.sample_rate().0 as f32 / 16000.0;
-         let mut buffer = vec![];
-         let mut last = Instant::now();
-         let stream = device.build_input_stream(
-            &cfg.clone().into(),
-            move |data: &[f32], _: &cpal::InputCallbackInfo| {
-               let mono_data: Vec<f32> = data.iter().map(|x| *x).step_by(channels.into()).collect();
-               let mut downsampled_data = vec![];
-               for i in 0..(mono_data.len() as f32 / downsample_ratio) as usize {
-                  let mut upsampled = i as f32 * downsample_ratio;
-                  if upsampled > (mono_data.len()-1) as f32 {
-                     upsampled = (mono_data.len()-1) as f32
-                  }
-                  downsampled_data.push(mono_data[upsampled as usize]);
-               }
-               downsampled_data.iter().for_each(|x| buffer.push(*x));
-               if Instant::now() - last > stream_step {
-                  let result = w.transcribe(&buffer).unwrap();
-                  eprintln!("{}", chrono::Local::now());
-                  println!("{}", result);
-
-                  let retain = buffer.len() - (buffer.len() as f32 * stream_churn) as usize;
-                  for i in retain..buffer.len() {
-                     buffer[i - retain] = buffer[i]
-                  }
-                  buffer.truncate(retain);
-                  last = Instant::now();
-               }
-            },
-            move |err| {
-               eprintln!("input error: {}", err)
-            },
-            None,
-         ).unwrap();
-         stream.play().unwrap();
-
-         eprintln!("listening on {}", device.name().unwrap());
-         let mut signals = Signals::new(&[SIGINT]).unwrap();
-         for sig in signals.forever() {
-            eprintln!("sig {}", sig);
-            break;
-         }
-         stream.pause().unwrap();
-      },
-   };
-}
-
-struct Whisper {
-   ctx: WhisperContext,
-   threads: i32,
-}
-
-fn new_whisper(model_path: String, threads: i32) -> Result<Whisper, String> {
-   match WhisperContext::new(&model_path) {
-      Ok(ctx) => Ok(Whisper{
-         ctx: ctx,
-         threads: threads,
-      }),
-      Err(msg) => Err(format!("failed to load {}: {}", model_path, msg)),
-   }
-}
-
-impl Whisper {
-   fn transcribe(&self, data: &Vec<f32>) -> Result<String, String> {
-      match self._transcribe(&data) {
-         Ok(result) => Ok(result),
-         Err(msg) => Err(format!("failed to transcribe: {}", msg)),
-      }
-   }
-
-   fn _transcribe(&self, data: &Vec<f32>) -> Result<String, WhisperError> {
-      let mut params = FullParams::new(SamplingStrategy::Greedy { best_of: 0 });
-      params.set_no_context(true);
-      params.set_n_threads(self.threads);
-      params.set_translate(false);
-      params.set_detect_language(false);
-      params.set_language(Some("en"));
-      params.set_print_special(false);
-      params.set_print_progress(false);
-      params.set_print_realtime(false);
-      params.set_print_timestamps(false);
-
-      let mut state = self.ctx.create_state()?;
-      state.full(params, &data[..])?;
-
-      let num_segments = state.full_n_segments()?;
-      let mut result = "".to_string();
-      for i in 0..num_segments {
-         let segment = state.full_get_segment_text(i)?;
-         result = format!("{} {}", result, segment);
-      }
-
-      Ok(result)
-   }
-}

snowboy-2022/build.sh

@@ -1,99 +0,0 @@
-#! /bin/bash
-
-echo https://github.com/seasalt-ai/snowboy
-cd "$(dirname "$(realpath "$BASH_SOURCE")")"
-
-set -e
-set -o pipefail
-
-if [ ! -d ./snowboy.git.d ]; then
-   git clone https://github.com/seasalt-ai/snowboy snowboy.git.d
-fi
-
-timeout 2 docker version &> /dev/null
-if ! docker images | grep snowboy-pmdl.*latest &> /dev/null; then
-   pushd snowboy.git.d
-   docker build -t snowboy-pmdl:latest .
-   popd
-fi
-
-export HOTWORD="${HOTWORD:-${TRAIN:-default_hotword}}"
-
-if [ -n "$TRAIN" ] || [ ! -d ./model ] || [ ! -f ./model/$HOTWORD.pmdl ]; then
-   mkdir -p model
-   pushd model
-   rm -f ./record{1,2,3}.wav || true
-   echo "record 3 instances of '$HOTWORD'" >&2
-   for i in 1 2 3; do
-      read -p "[$i/3] ready? you get 3 seconds."
-      (
-         timeout 3 rec \
-            -r 16000 \
-            -c 1 \
-            -b 16 \
-            -e signed-integer \
-            -t wav \
-            record$i.wav
-      ) || true
-      ls record$i.wav
-   done
-   popd
-
-   docker run \
-      --rm \
-      -it \
-      -v "$(realpath ./model)":/snowboy-master/examples/Python/model \
-      snowboy-pmdl:latest
-   mv ./model/hotword.pmdl ./model/$HOTWORD.pmdl
-   if [ -n "$TRAIN" ]; then
-      exit 0
-   fi
-fi
-
-if false; then
-   if ! which swig; then
-      brew install swig
-   fi
-   pip3 install pyaudio
-   pushd snowboy.git.d/swig/Python3/
-   make
-   popd
-
-   cd snowboy.git.d/examples/Python3/
-   echo '
-   import snowboydecoder
-   import datetime
-   detected_callback = lambda *args: print(datetime.datetime.now(), "GOTCHA")
-   d = snowboydecoder.HotwordDetector("../../../model/'"$HOTWORD"'.pmdl", sensitivity=0.5, audio_gain=1)
-   d.start(detected_callback)
-   ' > breel.py
-   echo GO
-   cleanup() {
-      echo OK IM DONE NOW
-   }
-   trap cleanup EXIT
-   python3 ./breel.py
-else
-   resources="$(realpath snowboy.git.d/resources/common.res)"
-   hotword="$(realpath ./model/$HOTWORD.pmdl)"
-   GOPROXY= go build -o snowboy
-   if [ -z "$PUSH" ]; then
-      ./snowboy \
-         -ms "$hotword/$HOTWORD" \
-         -r "$resources" \
-         -s 0.5 \
-         "$@"
-   else
-      echo '
-FROM registry-app.eng.qops.net:5001/imported/alpine:3.16
-WORKDIR /main/
-COPY ./snowboy.git.d/resources/common.res ./
-COPY ./model/hotword.pmdl ./
-COPY ./snowboy ./
-ENTRYPOINT ["sh", "-c", "true; echo copying /main/ to /mnt/; cp /main/* /mnt/"]
-CMD []
-      ' > Dockerfile
-      docker build -t registry-app.eng.qops.net:5001/breel/snowboy:latest .
-      docker push registry-app.eng.qops.net:5001/breel/snowboy:latest
-   fi
-fi

snowboy-2022/go.mod

@@ -1,13 +0,0 @@
-module snowboy
-
-go 1.19
-
-require (
-	github.com/brentnd/go-snowboy v0.0.0-20190301212623-e19133c572af
-	github.com/gordonklaus/portaudio v0.0.0-20221027163845-7c3b689db3cc
-)
-
-require (
-	github.com/Kitt-AI/snowboy v1.3.0 // indirect
-	github.com/stretchr/testify v1.8.1 // indirect
-)

snowboy-2022/go.sum

@@ -1,22 +0,0 @@
-github.com/Kitt-AI/snowboy v1.3.0 h1:PjBVN84M/9tAzDBQXILAKMoJMxt/fT0nhJ1rhKtVRUc=
-github.com/Kitt-AI/snowboy v1.3.0/go.mod h1:sDzzMXFQ1wFkXkZaX/ant0xJsizGVq/9hyKb7ZB3cNI=
-github.com/brentnd/go-snowboy v0.0.0-20190301212623-e19133c572af h1:ijY5OHNQs3CdzTN2XT+zByIsR1QVyXTvOUSkQcBm6pw=
-github.com/brentnd/go-snowboy v0.0.0-20190301212623-e19133c572af/go.mod h1:XcT4k8Tn9hrM5SLVvu5hNQbAC6GojXM0MXz1Rt8CL68=
-github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
-github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
-github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
-github.com/gordonklaus/portaudio v0.0.0-20221027163845-7c3b689db3cc h1:yYLpN7bJxKYILKnk20oczGQOQd2h3/7z7/cxdD9Se/I=
-github.com/gordonklaus/portaudio v0.0.0-20221027163845-7c3b689db3cc/go.mod h1:WY8R6YKlI2ZI3UyzFk7P6yGSuS+hFwNtEzrexRyD7Es=
-github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
-github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
-github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
-github.com/stretchr/objx v0.4.0/go.mod h1:YvHI0jy2hoMjB+UWwv71VJQ9isScKT/TqJzVSSt89Yw=
-github.com/stretchr/objx v0.5.0/go.mod h1:Yh+to48EsGEfYuaHDzXPcE3xhTkx73EhmCGUpEOglKo=
-github.com/stretchr/testify v1.7.1/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
-github.com/stretchr/testify v1.8.0/go.mod h1:yNjHg4UonilssWZ8iaSj1OCr/vHnekPRkoO+kdMU+MU=
-github.com/stretchr/testify v1.8.1 h1:w7B6lhMri9wdJUVmEZPGGhZzrYTPvgJArz7wNPgYKsk=
-github.com/stretchr/testify v1.8.1/go.mod h1:w2LPCIKwWwSfY2zedu0+kehJoqGctiVI29o6fzry7u4=
-gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
-gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
-gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
-gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=

snowboy-2022/main.go

@@ -1,122 +0,0 @@
-// This example streams the microphone thru Snowboy to listen for the hotword,
-// by using the PortAudio interface.
-//
-// HOW TO USE:
-// 	go run examples/Go/listen/main.go [path to snowboy resource file] [path to snowboy hotword file]
-//
-package main
-
-import (
-	"bytes"
-	"encoding/binary"
-	"flag"
-	"fmt"
-	"log"
-	"path"
-	"strings"
-	"time"
-
-	"github.com/brentnd/go-snowboy"
-	"github.com/gordonklaus/portaudio"
-)
-
-// Sound represents a sound stream implementing the io.Reader interface
-// that provides the microphone data.
-type Sound struct {
-	stream *portaudio.Stream
-	data   []int16
-}
-
-// Init initializes the Sound's PortAudio stream.
-func (s *Sound) Init() {
-	inputChannels := 1
-	outputChannels := 0
-	sampleRate := 16000
-	s.data = make([]int16, 1024)
-
-	// initialize the audio recording interface
-	err := portaudio.Initialize()
-	if err != nil {
-		fmt.Errorf("Error initialize audio interface: %s", err)
-		return
-	}
-
-	// open the sound input stream for the microphone
-	stream, err := portaudio.OpenDefaultStream(inputChannels, outputChannels, float64(sampleRate), len(s.data), s.data)
-	if err != nil {
-		fmt.Errorf("Error open default audio stream: %s", err)
-		return
-	}
-
-	err = stream.Start()
-	if err != nil {
-		fmt.Errorf("Error on stream start: %s", err)
-		return
-	}
-
-	s.stream = stream
-}
-
-// Close closes down the Sound's PortAudio connection.
-func (s *Sound) Close() {
-	s.stream.Close()
-	portaudio.Terminate()
-}
-
-// Read is the Sound's implementation of the io.Reader interface.
-func (s *Sound) Read(p []byte) (int, error) {
-	s.stream.Read()
-
-	buf := &bytes.Buffer{}
-	for _, v := range s.data {
-		binary.Write(buf, binary.LittleEndian, v)
-	}
-
-	copy(p, buf.Bytes())
-	return len(p), nil
-}
-
-func main() {
-	resources := flag.String("r", "", "path to the .res file")
-	models := flag.String("ms", "", "comma delimited path to the .?mdl file/output")
-	sensitivity := flag.Float64("s", 0.45, "0..1")
-	quiet := flag.Bool("q", false, "emit '1' on detect else silent")
-	flag.Parse()
-
-	if *resources == "" || *models == "" {
-		panic("all flags must be set")
-	}
-
-	// open the mic
-	mic := &Sound{}
-	mic.Init()
-	defer mic.Close()
-
-	// open the snowboy detector
-	d := snowboy.NewDetector(*resources)
-	defer d.Close()
-
-	// set the handlers
-	for _, modelStrC := range strings.Split(*models, ",") {
-		modelStr := modelStrC
-		d.HandleFunc(snowboy.NewHotword(path.Dir(modelStr), float32(*sensitivity)), func(string) {
-			if !*quiet {
-				log.Println(path.Base(modelStr))
-			}
-			fmt.Println(path.Base(modelStr))
-		})
-	}
-
-	d.HandleSilenceFunc(1*time.Second, func(string) {
-		if !*quiet {
-			log.Println("...")
-		}
-	})
-
-	// display the detector's expected audio format
-	sr, nc, bd := d.AudioFormat()
-	log.Printf("sample rate=%d, num channels=%d, bit depth=%d\n", sr, nc, bd)
-
-	// start detecting using the microphone
-	d.ReadAndDetect(mic)
-}

snowboy-2022/pyautogui.sh

@@ -1,82 +0,0 @@
-#! /bin/bash
-
-main() {
-   cleanup() {
-      killall -9 $(jobs -p)
-      killall snowboy
-   }
-   trap cleanup EXIT
-
-   if [ ! -e /tmp/stt.fifo ]; then
-      mkfifo /tmp/stt.fifo
-   fi
-
-   echo starting in
-   for ((i=2; i>0; i--)); do
-      echo "...$i..."
-      sleep 1
-   done
-
-   local models=($(
-      cat pyautogui.yaml \
-         | gojq -r -c --yaml-input '
-            to_entries[] | "model/"+.key+".pmdl/"+.key
-         ' \
-      | tr '\n' ',' \
-      | sed 's/,$//'
-   ))
-   echo models=$models
-
-   ./snowboy -r resources.res -ms $models "$@" > /tmp/stt.fifo &
-
-   python3 -c '
-import pyautogui
-import time
-
-keys = set()
-
-def toggle(key):
-   global keys
-   if key in keys:
-      release(key)
-   else:
-      hold(key)
-
-def hold(key):
-   global keys
-   for keyin in [todrop for todrop in keys]:
-      if keyin != key:
-         release(keyin)
-   keys = set()
-   keys.add(key)
-   print()
-   print("pressing", key)
-   print()
-   pyautogui.keyDown(key)
-
-def release(key):
-   print()
-   print("releasing", key)
-   print()
-   pyautogui.keyUp(key)
-
-def main():
-   with open("/tmp/stt.fifo", "r") as q:
-      for line in q:
-         handle(line.strip())
-
-import yaml
-mapping = yaml.safe_load(open("./pyautogui.yaml", "r"))
-print(mapping)
-
-def handle(cmd):
-   global mapping
-   hold(mapping.get(cmd))
-
-main()
-   '
-}
-
-if [ "$0" == "$BASH_SOURCE" ]; then
-   main "$@"
-fi

snowboy-2022/pyautogui.yaml

@@ -1,5 +0,0 @@
-up: w
-down: s
-left: a
-right: d
-jump: w

src/main.rs

@@ -0,0 +1,374 @@
+use whisper_rs::{WhisperContext, FullParams, SamplingStrategy, WhisperError};
+use cpal::traits::{HostTrait, DeviceTrait, StreamTrait};
+use signal_hook::{iterator::Signals, consts::signal::SIGINT};
+use std::time::{Duration, Instant};
+use chrono;
+use clap::Parser;
+use std::thread;
+use std::fs::File;
+use std::io::Write;
+
+#[derive(Parser, Debug)]
+struct Flags {
+   #[arg(long, default_value = "./models/ggml-tiny.en.bin")]
+   model: String,
+
+   #[arg(long, default_value = "8")]
+   threads: i32,
+
+   #[arg(long, default_value = "5")]
+   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>,
+
+   #[arg(long, default_value = "false")]
+   debug: bool,
+}
+
+fn main() {
+   let flags = Flags::parse();
+
+   let w = new_whisper_service(
+      flags.model,
+      flags.threads,
+      flags.stream_head,
+      flags.stream_tail,
+      |result: Result<Whispered, String>| {
+         match result {
+            Ok(whispered) => {
+               eprintln!("{}: {:?}", chrono::Local::now(), whispered);
+               println!("{}", whispered.to_string());
+            },
+            Err(msg) => { eprintln!("Error whispering: {}", msg); },
+         };
+      },
+   ).unwrap();
+   let stream_retain = (flags.stream_retain * 16_000.0) as usize;
+   let stream_step = Duration::new(flags.stream_step, 0);
+   match flags.wav {
+      Some(wav) => {
+         let (header, data) = wav::read(
+            &mut std::fs::File::open(wav).expect("failed to open $WAV"),
+         ).expect("failed to decode $WAV");
+         assert!(header.channel_count == 1);
+         assert!(header.sampling_rate == 16_000);
+         let data16 = data.as_sixteen().expect("wav is not 32bit floats");
+         let audio_data = &whisper_rs::convert_integer_to_float_audio(&data16);
+
+         w.transcribe(&audio_data);
+      },
+      None => {
+         match &flags.debug {
+            true => { File::create("/tmp/page.rawf32audio").unwrap(); },
+            false => {},
+         };
+         let mut buffer = vec![];
+         let mut last = Instant::now();
+         new_listener().listen(move |data: Vec<f32>| {
+            data.iter().for_each(|x| buffer.push(*x));
+            if Instant::now() - last > stream_step {
+               w.transcribe_async(&buffer).unwrap();
+
+               match &flags.debug {
+                  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);
+               last = Instant::now();
+            }
+         });
+      },
+   };
+}
+
+struct WhisperService {
+   jobs: std::sync::mpsc::SyncSender<AWhisper>,
+}
+
+fn new_whisper_service<F>(model_path: String, threads: i32, stream_head: f32, stream_tail: f32, handler_fn: F) -> Result<WhisperService, String> where F: FnMut(Result<Whispered, String>) + Send + 'static {
+   match new_whisper_engine(model_path, threads) {
+      Ok(engine) => {
+         let mut whisper = new_whisper_impl(engine, stream_head, stream_tail, handler_fn);
+         let (send, recv) = std::sync::mpsc::sync_channel(100);
+         thread::spawn(move || { whisper.transcribe_asyncs(recv); });
+         Ok(WhisperService{jobs: send})
+      },
+      Err(msg) => Err(format!("failed to initialize engine: {}", msg)),
+   }
+}
+
+impl WhisperService {
+   fn transcribe(&self, data: &Vec<f32>) {
+      let (send, recv) = std::sync::mpsc::sync_channel(0);
+      self._transcribe_async(data, Some(send)).unwrap();
+      recv.recv().unwrap();
+   }
+
+   fn transcribe_async(&self, data: &Vec<f32>) -> Result<(), String> {
+      self._transcribe_async(data, None)
+   }
+
+   fn _transcribe_async(&self, data: &Vec<f32>, ack: Option<std::sync::mpsc::SyncSender<bool>>) -> Result<(), String> {
+      match self.jobs.try_send(AWhisper{
+         data: data.clone().to_vec(),
+         ack: ack,
+      }) {
+         Ok(_) => Ok(()),
+         Err(msg) => Err(format!("failed to enqueue transcription: {}", msg)),
+      }
+   }
+}
+
+struct WhisperImpl {
+   engine: WhisperEngine,
+   stream_head: f32,
+   stream_tail: f32,
+   handler_fn: Option<Box<dyn FnMut(Result<Whispered, String>) + Send + 'static>>
+}
+
+fn new_whisper_impl<F>(engine: WhisperEngine, stream_head: f32, stream_tail: f32, handler_fn: F) -> WhisperImpl where F: FnMut(Result<Whispered, String>) + Send + 'static {
+   WhisperImpl {
+      engine: engine,
+      stream_head: stream_head,
+      stream_tail: stream_tail,
+      handler_fn: Some(Box::new(handler_fn)),
+   }
+}
+
+impl WhisperImpl {
+   fn transcribe_asyncs(&mut self, recv: std::sync::mpsc::Receiver<AWhisper>) {
+      loop {
+         match recv.recv() {
+            Ok(job) => {
+               let result = self.transcribe(&job).is_ok();
+               match job.ack {
+                  Some(ack) => {
+                     ack.send(result).unwrap();
+                  },
+                  None => (),
+               };
+            }
+            Err(_) => return,
+         };
+      }
+   }
+
+   fn transcribe(&mut self, a_whisper: &AWhisper) -> Result<(), ()> {
+      match self.engine.transcribe(&a_whisper.data) {
+         Ok(result) => {
+            self.on_success(&result);
+            Ok(())
+         },
+         Err(msg) => {
+            self.on_error(msg.to_string());
+            Err(())
+         },
+      }
+   }
+
+   fn on_success(&mut self, whispered: &Whispered) {
+      let result = whispered
+         .after(&(self.stream_head * 100.0))
+         .before(&(self.stream_tail * 100.0));
+      (self.handler_fn.as_mut().unwrap())(Ok(result));
+   }
+
+   fn on_error(&mut self, msg: String) {
+      (self.handler_fn.as_mut().unwrap())(Err(format!("failed to transcribe: {}", &msg)));
+   }
+}
+
+struct WhisperEngine {
+   ctx: WhisperContext,
+   threads: i32,
+}
+
+fn new_whisper_engine(model_path: String, threads: i32) -> Result<WhisperEngine, String> {
+   match WhisperContext::new(&model_path) {
+      Ok(ctx) => Ok(WhisperEngine{ctx: ctx, threads: threads}),
+      Err(msg) => Err(format!("failed to load {}: {}", model_path, msg)),
+   }
+}
+
+impl WhisperEngine {
+   fn transcribe(&self, data: &Vec<f32>) -> Result<Whispered, WhisperError> {
+      let mut params = FullParams::new(SamplingStrategy::Greedy { best_of: 0 });
+      params.set_no_context(true);
+      params.set_n_threads(self.threads);
+      params.set_translate(false);
+      params.set_detect_language(false);
+      params.set_language(Some("en"));
+      params.set_print_special(false);
+      params.set_print_progress(false);
+      params.set_print_realtime(false);
+      params.set_print_timestamps(false);
+
+      let mut state = self.ctx.create_state()?;
+      state.full(params, &data[..])?;
+
+      let mut result = new_whispered();
+      let num_segments = state.full_n_segments()?;
+      for i in 0..num_segments {
+         let data = state.full_get_segment_text(i)?;
+         let start = state.full_get_segment_t0(i)?;
+         let stop = state.full_get_segment_t1(i)?;
+         result.push(data, start, stop);
+      }
+
+      Ok(result)
+   }
+}
+
+struct AWhisper {
+   data: Vec<f32>,
+   ack: Option<std::sync::mpsc::SyncSender<bool>>,
+}
+
+#[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 Listener {
+}
+
+fn new_listener() -> Listener {
+   Listener{}
+}
+
+impl Listener {
+   fn listen(self, mut cb: impl FnMut(Vec<f32>)) {
+      let (send, recv) = std::sync::mpsc::sync_channel(100);
+      thread::spawn(move || { self._listen(send); });
+      loop {
+         match recv.recv() {
+            Ok(msg) => cb(msg),
+            Err(_) => return,
+         };
+      }
+   }
+
+   fn _listen(self, send: std::sync::mpsc::SyncSender<Vec<f32>>) {
+      let host = cpal::default_host();
+      let device = host.default_input_device().unwrap();
+      let cfg = device.supported_input_configs()
+         .unwrap()
+         .filter(|x| x.sample_format() == cpal::SampleFormat::F32)
+         .nth(0)
+         .unwrap()
+         .with_max_sample_rate();
+
+      let downsample_ratio = cfg.channels() as f32 * (cfg.sample_rate().0 as f32 / 16_000.0);
+      let stream = device.build_input_stream(
+         &cfg.clone().into(),
+         move |data: &[f32], _: &cpal::InputCallbackInfo| {
+            let mut downsampled_data = vec![];
+            for i in 0..(data.len() as f32 / downsample_ratio) as usize {
+               let mut upsampled = i as f32 * downsample_ratio;
+               if upsampled > (data.len()-1) as f32 {
+                  upsampled = (data.len()-1) as f32
+               }
+               downsampled_data.push(data[upsampled as usize]);
+            }
+            match send.try_send(downsampled_data) {
+               Ok(_) => (),
+               Err(msg) => eprintln!("failed to ingest audio: {}", msg),
+            };
+         },
+         move |err| {
+            eprintln!("input error: {}", err)
+         },
+         None,
+      ).unwrap();
+      stream.play().unwrap();
+
+      eprintln!("listening on {}", device.name().unwrap());
+      let mut signals = Signals::new(&[SIGINT]).unwrap();
+      for sig in signals.forever() {
+         eprintln!("sig {}", sig);
+         break;
+      }
+      stream.pause().unwrap();
+   }
+}