mirror of
https://github.com/Smaug123/PulumiConfig
synced 2025-10-18 14:28:40 +00:00
Whisper (#26)
This commit is contained in:
Patrick Stevens
GitHub
2
PulumiWebServer/Nix/whisper/requirements.txt
Normal file
2
PulumiWebServer/Nix/whisper/requirements.txt
Normal file
@@ -0,0 +1,2 @@
|
||||
flask
|
||||
waitress
|
||||
107
PulumiWebServer/Nix/whisper/transcribe.html
Normal file
107
PulumiWebServer/Nix/whisper/transcribe.html
Normal file
@@ -0,0 +1,107 @@
|
||||
<!DOCTYPE html>
|
||||
<html lang="en">
|
||||
<head>
|
||||
<title>Whisper Transcription</title>
|
||||
<style>
|
||||
#output {
|
||||
white-space: pre-wrap;
|
||||
border: 1px solid #ccc;
|
||||
padding: 10px;
|
||||
margin: 10px;
|
||||
width: 95%;
|
||||
height: 300px;
|
||||
overflow-y: auto;
|
||||
}
|
||||
</style>
|
||||
</head>
|
||||
<body>
|
||||
<h1>Whisper Transcription</h1>
|
||||
<p>Submit file for transcription</p>
|
||||
<form action="/upload" method="POST" enctype="multipart/form-data">
|
||||
<input type="file" name="file">
|
||||
<input type="submit" value="Submit">
|
||||
</form>
|
||||
|
||||
<div>
|
||||
<label for="file-to-analyze">File to analyze:</label>
|
||||
<div contenteditable="true" id="file-to-analyze">{no file set}</div>
|
||||
</div>
|
||||
<button id="start">Start analysing</button>
|
||||
|
||||
<button id="displayWav" hidden="hidden">Listen to file being transcribed</button>
|
||||
<div id="wavContainer"></div>
|
||||
|
||||
<div id="status"></div>
|
||||
<div id="output"></div>
|
||||
|
||||
<script>
|
||||
const uploadForm = document.querySelector('form');
|
||||
const uploadResultDiv = document.getElementById('file-to-analyze');
|
||||
uploadForm.addEventListener('submit', e => {
|
||||
e.preventDefault();
|
||||
|
||||
const files = document.querySelector('[type=file]').files;
|
||||
const formData = new FormData();
|
||||
formData.append('file', files[0]);
|
||||
|
||||
fetch('/upload', {
|
||||
method: 'POST',
|
||||
body: formData
|
||||
})
|
||||
.then(response => response.text())
|
||||
.then(response => {
|
||||
uploadResultDiv.innerText = response;
|
||||
})
|
||||
.catch(error => console.error(error))
|
||||
});
|
||||
|
||||
const outputDiv = document.getElementById('output');
|
||||
const statusDiv = document.getElementById('status');
|
||||
const displayButton = document.getElementById('displayWav');
|
||||
const wavContainer = document.getElementById('wavContainer');
|
||||
const startButton = document.getElementById('start');
|
||||
|
||||
startButton.onclick = function() {
|
||||
// Create a new EventSource instance pointing to the SSE route
|
||||
// const eventSource = new EventSource('/transcribe-youtube?url=https://www.youtube.com/watch?v=-xZQ0YZ7ls4');
|
||||
const eventSource = new EventSource('/transcribe-file?file=' + uploadResultDiv.innerText);
|
||||
|
||||
let file = '';
|
||||
|
||||
displayButton.onclick = function () {
|
||||
const audioElt = document.createElement('audio');
|
||||
audioElt.controls = true;
|
||||
audioElt.src = '/download?file=' + file;
|
||||
|
||||
wavContainer.innerHTML = '';
|
||||
wavContainer.appendChild(audioElt);
|
||||
};
|
||||
|
||||
eventSource.addEventListener('started', function (e) {
|
||||
statusDiv.innerText = 'Transcription has begun. Please hold the line; my server is only very small and weedy.';
|
||||
displayButton.hidden = false;
|
||||
file = e.data;
|
||||
});
|
||||
|
||||
eventSource.addEventListener('quit', function (e) {
|
||||
statusDiv.innerText = 'Transcription finished';
|
||||
eventSource.close()
|
||||
});
|
||||
|
||||
eventSource.onmessage = function (e) {
|
||||
outputDiv.innerText += e.data + '\n';
|
||||
};
|
||||
|
||||
// Handle any errors
|
||||
eventSource.onerror = function (e) {
|
||||
if (eventSource.readyState === EventSource.CLOSED) {
|
||||
console.log('Connection was closed');
|
||||
} else {
|
||||
outputDiv.innerText += 'Error! Connection was lost. Refresh the page to retry.\n';
|
||||
eventSource.close()
|
||||
}
|
||||
};
|
||||
};
|
||||
</script>
|
||||
</body>
|
||||
</html>
|
||||
65
PulumiWebServer/Nix/whisper/whisper.nix
Normal file
65
PulumiWebServer/Nix/whisper/whisper.nix
Normal file
@@ -0,0 +1,65 @@
|
||||
{
|
||||
config,
|
||||
pkgs,
|
||||
lib,
|
||||
whisper-packages,
|
||||
...
|
||||
}: {
|
||||
options = {
|
||||
services.whisper-config = {
|
||||
domain = lib.mkOption {
|
||||
type = lib.types.str;
|
||||
example = "example.com";
|
||||
description = lib.mdDoc "Top-level domain to configure";
|
||||
};
|
||||
subdomain = lib.mkOption {
|
||||
type = lib.types.str;
|
||||
example = "whisper";
|
||||
description = lib.mdDoc "Subdomain in which to put the Whisper server";
|
||||
};
|
||||
port = lib.mkOption {
|
||||
type = lib.types.port;
|
||||
description = lib.mdDoc "Whisper localhost port to be forwarded";
|
||||
default = 1739;
|
||||
};
|
||||
};
|
||||
};
|
||||
|
||||
config = {
|
||||
users.users."whisper".extraGroups = [config.users.groups.keys.name];
|
||||
users.users."whisper".group = "whisper";
|
||||
users.groups.whisper = {};
|
||||
users.users."whisper".isSystemUser = true;
|
||||
|
||||
systemd.services.whisper-server = {
|
||||
description = "whisper-server";
|
||||
wantedBy = ["multi-user.target"];
|
||||
serviceConfig = let
|
||||
python = pkgs.python3.withPackages (p: with p; [flask waitress]);
|
||||
in {
|
||||
Restart = "always";
|
||||
Type = "exec";
|
||||
User = "whisper";
|
||||
Group = "whisper";
|
||||
ExecStart = "${python}/bin/python ${./whisper.py}";
|
||||
};
|
||||
environment = {
|
||||
WHISPER_NORMALIZE = "${whisper-packages.normalize}/bin/normalize.sh";
|
||||
WHISPER_CLIENT = "${whisper-packages.default}/bin/whisper-cpp";
|
||||
WHISPER_PORT = toString config.services.whisper-config.port;
|
||||
INDEX_PAGE_PATH = ./transcribe.html;
|
||||
YT_DLP = "${pkgs.yt-dlp}/bin/yt-dlp";
|
||||
};
|
||||
};
|
||||
|
||||
services.nginx.proxyTimeout = "300s";
|
||||
services.nginx.clientMaxBodySize = "50M";
|
||||
services.nginx.virtualHosts."${config.services.whisper-config.subdomain}.${config.services.whisper-config.domain}" = {
|
||||
forceSSL = true;
|
||||
enableACME = true;
|
||||
locations."/" = {
|
||||
proxyPass = "http://localhost:${toString config.services.whisper-config.port}/";
|
||||
};
|
||||
};
|
||||
};
|
||||
}
|
||||
129
PulumiWebServer/Nix/whisper/whisper.py
Normal file
129
PulumiWebServer/Nix/whisper/whisper.py
Normal file
@@ -0,0 +1,129 @@
|
||||
import subprocess
|
||||
import os
|
||||
from typing import AnyStr
|
||||
import re
|
||||
from flask import Flask, Response, request, render_template_string
|
||||
import waitress
|
||||
import tempfile
|
||||
|
||||
app = Flask(__name__)
|
||||
|
||||
youtube_regex = re.compile(
|
||||
r"^(?:https?://)?(?:www\.)?(?:youtu\.be/|youtube\.com/(?:embed/|v/|watch\?v=|watch\?.+&v=))((\w|-){11})(?:\S+)?$")
|
||||
|
||||
alnum_regex = re.compile(r"^[a-zA-Z0-9]+$")
|
||||
|
||||
|
||||
def generate_output(wav_file):
|
||||
process = subprocess.Popen([whisper, "--file", f"/tmp/whisper/{wav_file}.wav", "--output-txt"],
|
||||
stdout=subprocess.PIPE, bufsize=1,
|
||||
text=True)
|
||||
|
||||
yield f'event: started\ndata: {wav_file}\n\n'
|
||||
|
||||
for line in iter(process.stdout.readline, ''):
|
||||
yield f"data: {line}\n\n"
|
||||
|
||||
yield 'event: quit\ndata: \n\n'
|
||||
|
||||
os.remove(f"/tmp/whisper/{wav_file}.wav")
|
||||
|
||||
|
||||
def obtain_youtube(url: AnyStr) -> str:
|
||||
# handle, temp_file = tempfile.mkstemp(".wav", text=False)
|
||||
# os.close(handle)
|
||||
# os.remove(temp_file)
|
||||
|
||||
# output = subprocess.run(
|
||||
# [ytdlp, '--extract-audio', '--audio-format', 'wav', '--cookies', '/tmp/cookies.txt', '--audio-quality', '16k', '--force-ipv6', '--output', temp_file,
|
||||
# url], check=True, capture_output=True, text=True)
|
||||
# if "429 Too Many Requests" in output.stdout:
|
||||
# raise subprocess.CalledProcessError(1, whisper, "YouTube replied saying Too Many Requests")
|
||||
# return temp_file
|
||||
|
||||
raise Exception("DigitalOcean is rate limited to YouTube")
|
||||
|
||||
|
||||
def normalize(path: str, output: str):
|
||||
try:
|
||||
subprocess.run([normalize_binary, path, output], check=True)
|
||||
except subprocess.CalledProcessError:
|
||||
os.remove(path)
|
||||
return Response("failed to normalize", status=500)
|
||||
|
||||
|
||||
@app.route('/transcribe-youtube')
|
||||
def transcribe_youtube():
|
||||
try:
|
||||
url = request.args.get('url')
|
||||
except KeyError:
|
||||
return Response("must have a URL in the format ?url=https://www.youtube.com/watch?v=...", status=400)
|
||||
if youtube_regex.match(url) is None:
|
||||
return Response(f"url '{url}' did not appear to be a YouTube video", status=400)
|
||||
wav_file = obtain_youtube(url)
|
||||
return Response(generate_output(wav_file), mimetype="text/event-stream")
|
||||
|
||||
|
||||
@app.route('/transcribe-file')
|
||||
def transcribe_file():
|
||||
try:
|
||||
file = request.args.get('file')
|
||||
except KeyError:
|
||||
return Response("must have a file as obtained from /upload, in the format ?file=...", status=400)
|
||||
if alnum_regex.match(file) is None:
|
||||
return Response(f"filename '{file}' was not alphanumeric", status=400)
|
||||
return Response(generate_output(file), mimetype="text/event-stream")
|
||||
|
||||
|
||||
@app.route('/transcribe-ui')
|
||||
def index():
|
||||
return render_template_string(open(index_page_path).read()) # Assuming 'index.html' is in the same directory
|
||||
|
||||
|
||||
@app.route('/upload', methods=["POST"])
|
||||
def upload():
|
||||
if 'file' not in request.files:
|
||||
return 'No "file" part in request', 400
|
||||
file = request.files['file']
|
||||
|
||||
# Create temp file for this upload
|
||||
handle, temp_file = tempfile.mkstemp(text=False)
|
||||
try:
|
||||
os.close(handle)
|
||||
file.save(temp_file)
|
||||
# get filename from absolute path
|
||||
temp_file_frag = os.path.basename(temp_file)
|
||||
|
||||
normalize(temp_file, f"/tmp/whisper/{temp_file_frag}")
|
||||
finally:
|
||||
try:
|
||||
os.remove(temp_file)
|
||||
finally:
|
||||
pass
|
||||
|
||||
return Response(temp_file_frag, mimetype="text/plain")
|
||||
|
||||
|
||||
@app.route('/download')
|
||||
def download():
|
||||
try:
|
||||
file = request.args.get('file')
|
||||
except KeyError:
|
||||
return Response("must have a file parameter", status=400)
|
||||
|
||||
if alnum_regex.match(file) is None:
|
||||
return Response(f"file '{file}' was not alphanumeric, bad format", status=400)
|
||||
|
||||
return Response(open(f"/tmp/whisper/{file}.wav", 'rb').read(), mimetype="audio/wav")
|
||||
|
||||
|
||||
def run(port: int):
|
||||
waitress.serve(app, host="0.0.0.0", port=port)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
normalize_binary = os.environ["WHISPER_NORMALIZE"]
|
||||
whisper = os.environ["WHISPER_CLIENT"]
|
||||
index_page_path = os.environ["INDEX_PAGE_PATH"]
|
||||
ytdlp = os.environ["YT_DLP"]
|
||||
run(int(os.environ["WHISPER_PORT"]))
|
||||
Reference in New Issue
Block a user