Merge ec6c7ba25a into 8a2563ecab

2026-02-07 00:04:42 +01:00 · 2025-05-30 04:18:51 +07:00 · 2025-05-30 04:18:51 +07:00 · b710eba963
parent 8a2563ecab ec6c7ba25a
commit b710eba963
9 changed files with 341 additions and 24 deletions
--- a/openui_app.py
+++ b/openui_app.py
@ -0,0 +1,337 @@
+from flask import Flask, request, render_template_string, send_file, jsonify
+import os
+import uuid
+import threading
+
+app = Flask(__name__)
+
+UPLOAD_FOLDER = 'uploads'
+OUTPUT_FOLDER = 'outputs'
+os.makedirs(UPLOAD_FOLDER, exist_ok=True)
+os.makedirs(OUTPUT_FOLDER, exist_ok=True)
+
+# Global job storage
+jobs = {}
+processes = {}  # Track running processes for stop functionality
+
+HTML = '''
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <title>Tortoise TTS OpenUI</title>
+    <script src="https://cdn.tailwindcss.com"></script>
+</head>
+<body class="bg-gray-100 min-h-screen">
+<div class="container mx-auto max-w-6xl py-8">
+    <h1 class="text-3xl font-bold mb-6 text-center">Tortoise TTS OpenUI</h1>
+    <form id="tts-form" method="post" enctype="multipart/form-data" class="bg-white p-6 rounded-lg shadow-md">
+        <div class="grid grid-cols-1 md:grid-cols-3 gap-6">
+            <div class="space-y-4 col-span-1">
+                <!-- Left column: text and file upload -->
+                <label for="text" class="block font-semibold">Text to Synthesize:</label>
+                <textarea name="text" id="text" required class="w-full border rounded p-2" rows="6">{{ request.form.text or '' }}</textarea>
+                <label for="text_file" class="block font-semibold">Or upload a text file:</label>
+                <input type="file" name="text_file" id="text_file" accept=".txt" class="w-full border rounded p-2">
+            </div>
+            <div class="space-y-4 col-span-1">
+                <!-- Middle column: options -->
+                <label for="voice" class="block font-semibold">Voice (comma-separated, or 'random'):</label>
+                <input type="text" name="voice" id="voice" value="{{ request.form.voice or 'random' }}" class="w-full border rounded p-2">
+                <label for="preset" class="block font-semibold">Preset:</label>
+                <select name="preset" id="preset" class="w-full border rounded p-2">
+                    <option value="ultra_fast" {% if request.form.preset == 'ultra_fast' %}selected{% endif %}>Ultra Fast</option>
+                    <option value="fast" {% if request.form.preset == 'fast' or not request.form.preset %}selected{% endif %}>Fast</option>
+                    <option value="standard" {% if request.form.preset == 'standard' %}selected{% endif %}>Standard</option>
+                    <option value="high_quality" {% if request.form.preset == 'high_quality' %}selected{% endif %}>High Quality</option>
+                </select>
+                <label for="candidates" class="block font-semibold">Candidates:</label>
+                <input type="number" name="candidates" id="candidates" min="1" max="10" value="{{ request.form.candidates or 1 }}" class="w-full border rounded p-2">
+                <label for="seed" class="block font-semibold">Seed (optional):</label>
+                <input type="number" name="seed" id="seed" value="{{ request.form.seed or '' }}" class="w-full border rounded p-2">
+                <label for="device" class="block font-semibold">Device:</label>
+                <select name="device" id="device" class="w-full border rounded p-2">
+                    <option value="" {% if not request.form.device %}selected{% endif %}>Auto</option>
+                    <option value="cuda" {% if request.form.device == 'cuda' %}selected{% endif %}>CUDA (GPU)</option>
+                    <option value="cpu" {% if request.form.device == 'cpu' %}selected{% endif %}>CPU</option>
+                </select>
+                <label for="voices_dir" class="block font-semibold">Voices Dir (optional):</label>
+                <input type="text" name="voices_dir" id="voices_dir" value="{{ request.form.voices_dir or '' }}" class="w-full border rounded p-2">
+                <label for="text_split" class="block font-semibold">Text Split (e.g. 200,400):</label>
+                <input type="text" name="text_split" id="text_split" value="{{ request.form.text_split or '' }}" class="w-full border rounded p-2">
+                <label for="num_autoregressive_samples" class="block font-semibold">Num Autoregressive Samples:</label>
+                <input type="number" name="num_autoregressive_samples" id="num_autoregressive_samples" min="1" value="{{ request.form.num_autoregressive_samples or '' }}" class="w-full border rounded p-2">
+                <label for="temperature" class="block font-semibold">Temperature:</label>
+                <input type="number" step="any" name="temperature" id="temperature" value="{{ request.form.temperature or '' }}" class="w-full border rounded p-2">
+                <label for="length_penalty" class="block font-semibold">Length Penalty:</label>
+                <input type="number" step="any" name="length_penalty" id="length_penalty" value="{{ request.form.length_penalty or '' }}" class="w-full border rounded p-2">
+                <label for="repetition_penalty" class="block font-semibold">Repetition Penalty:</label>
+                <input type="number" step="any" name="repetition_penalty" id="repetition_penalty" value="{{ request.form.repetition_penalty or '' }}" class="w-full border rounded p-2">
+                <label for="top_p" class="block font-semibold">Top-p:</label>
+                <input type="number" step="any" name="top_p" id="top_p" value="{{ request.form.top_p or '' }}" class="w-full border rounded p-2">
+                <label for="max_mel_tokens" class="block font-semibold">Max Mel Tokens:</label>
+                <input type="number" name="max_mel_tokens" id="max_mel_tokens" min="1" max="600" value="{{ request.form.max_mel_tokens or '' }}" class="w-full border rounded p-2">
+                <label for="cvvp_amount" class="block font-semibold">CVVP Amount:</label>
+                <input type="number" step="any" name="cvvp_amount" id="cvvp_amount" value="{{ request.form.cvvp_amount or '' }}" class="w-full border rounded p-2">
+                <label for="diffusion_iterations" class="block font-semibold">Diffusion Iterations:</label>
+                <input type="number" name="diffusion_iterations" id="diffusion_iterations" value="{{ request.form.diffusion_iterations or '' }}" class="w-full border rounded p-2">
+                <label class="inline-flex items-center"><input type="checkbox" name="cond_free" {% if request.form.get('cond_free') %}checked{% endif %} class="mr-2"> Conditioning Free Diffusion</label>
+                <label for="cond_free_k" class="block font-semibold">Cond Free K:</label>
+                <input type="number" step="any" name="cond_free_k" id="cond_free_k" value="{{ request.form.cond_free_k or '' }}" class="w-full border rounded p-2">
+                <label for="diffusion_temperature" class="block font-semibold">Diffusion Temperature:</label>
+                <input type="number" step="any" name="diffusion_temperature" id="diffusion_temperature" value="{{ request.form.diffusion_temperature or '' }}" class="w-full border rounded p-2">
+                <div class="flex flex-wrap gap-4 mt-4">
+                    <label class="inline-flex items-center"><input type="checkbox" name="quiet" {% if request.form.get('quiet') %}checked{% endif %} class="mr-2"> Quiet</label>
+                    <label class="inline-flex items-center"><input type="checkbox" name="produce_debug_state" {% if request.form.get('produce_debug_state') %}checked{% endif %} class="mr-2"> Produce Debug State</label>
+                    <label class="inline-flex items-center"><input type="checkbox" name="skip_existing" {% if request.form.get('skip_existing') %}checked{% endif %} class="mr-2"> Skip Existing</label>
+                    <label class="inline-flex items-center"><input type="checkbox" name="disable_redaction" {% if request.form.get('disable_redaction') %}checked{% endif %} class="mr-2"> Disable Redaction</label>
+                </div>
+            </div>
+            <div class="flex flex-col col-span-1 space-y-4">
+                <!-- Right column: progress, log, synthesize/stop buttons, result -->
+                <div id="progress-section" class="mb-4 hidden">
+                    <div class="w-full bg-gray-200 rounded-full h-6 mb-4 overflow-hidden relative">
+                        <div id="progress-bar" class="h-6 bg-gradient-to-r from-blue-500 via-blue-400 to-blue-600 bg-[length:40px_40px] animate-stripes rounded-full absolute left-0 top-0 w-full"></div>
+                    </div>
+                    <div id="log-area" class="bg-black text-green-400 font-mono p-4 rounded-lg overflow-y-auto max-h-64 whitespace-pre-line shadow-inner"></div>
+                </div>
+                <div class="flex gap-4">
+                    <button type="submit" id="synthesize-btn" class="flex-1 bg-blue-600 text-white font-bold py-2 px-4 rounded hover:bg-blue-700 transition">Synthesize</button>
+                    <button type="button" id="stop-btn" class="flex-1 bg-red-600 text-white font-bold py-2 px-4 rounded hover:bg-red-700 transition hidden">Stop</button>
+                </div>
+                <div id="result-section" class="result bg-white p-6 rounded-lg shadow-md hidden">
+                    <h2 class="text-xl font-bold mb-2">Result</h2>
+                    <audio id="audio-player" controls class="w-full"></audio>
+                    <p class="mt-2"><a id="download-link" href="#" download class="text-blue-600 hover:underline">Download Audio</a></p>
+                </div>
+            </div>
+        </div>
+    </form>
+</div>
+<script>
+const form = document.getElementById('tts-form');
+const progressSection = document.getElementById('progress-section');
+const progressBar = document.getElementById('progress-bar');
+const logArea = document.getElementById('log-area');
+const resultSection = document.getElementById('result-section');
+const audioPlayer = document.getElementById('audio-player');
+const downloadLink = document.getElementById('download-link');
+const synthesizeBtn = document.getElementById('synthesize-btn');
+const stopBtn = document.getElementById('stop-btn');
+let polling = false;
+let currentJobId = null;
+
+form.addEventListener('submit', function(e) {
+    e.preventDefault();
+    progressSection.classList.remove('hidden');
+    resultSection.classList.add('hidden');
+    progressBar.style.width = '100%';
+    logArea.textContent = '';
+    synthesizeBtn.disabled = true;
+    stopBtn.classList.remove('hidden');
+    const formData = new FormData(form);
+    fetch('/', {
+        method: 'POST',
+        body: formData
+    })
+    .then(response => response.json())
+    .then(data => {
+        if (data.job_id) {
+            currentJobId = data.job_id;
+            pollProgress(data.job_id);
+        }
+    });
+});
+
+stopBtn.addEventListener('click', function() {
+    if (currentJobId) {
+        fetch(`/stop/${currentJobId}`, { method: 'POST' })
+            .then(() => {
+                polling = false;
+                stopBtn.classList.add('hidden');
+                synthesizeBtn.disabled = false;
+            });
+    }
+});
+
+function pollProgress(jobId) {
+    polling = true;
+    function poll() {
+        fetch(`/progress/${jobId}`)
+            .then(res => res.json())
+            .then(data => {
+                logArea.textContent = data.log;
+                logArea.scrollTop = logArea.scrollHeight;
+                if (data.done) {
+                    polling = false;
+                    stopBtn.classList.add('hidden');
+                    synthesizeBtn.disabled = false;
+                    if (data.audio_url) {
+                        audioPlayer.src = data.audio_url;
+                        downloadLink.href = data.audio_url;
+                        resultSection.classList.remove('hidden');
+                    }
+                } else if (polling) {
+                    setTimeout(poll, 500);
+                }
+            });
+    }
+    poll();
+}
+// Add Tailwind custom animation for stripes
+const style = document.createElement('style');
+style.innerHTML = `
+@keyframes stripes {
+  0% { background-position-x: 0; }
+  100% { background-position-x: 40px; }
+}
+.animate-stripes {
+  background-image: repeating-linear-gradient(135deg,rgba(255,255,255,0.15) 0 10px,transparent 10px 20px);
+  animation: stripes 1s linear infinite;
+}
+`;
+document.head.appendChild(style);
+</script>
+</body>
+</html>
+'''
+
+def run_tts_with_progress(cmd, env, job_id):
+    import subprocess
+    import re
+    jobs[job_id]['progress'] = 0
+    jobs[job_id]['log'] = ''
+    jobs[job_id]['done'] = False
+    process = subprocess.Popen(cmd, stdout=subprocess.PIPE, stderr=subprocess.STDOUT, env=env, text=True, bufsize=1)
+    processes[job_id] = process
+    try:
+        for line in process.stdout:
+            jobs[job_id]['log'] += line
+            match = re.search(r"\((\d+) of (\d+)\)", line)
+            if match:
+                current, total_clips = int(match.group(1)), int(match.group(2))
+                progress = int(current / total_clips * 100)
+                jobs[job_id]['progress'] = progress
+        process.wait()
+    finally:
+        jobs[job_id]['progress'] = 100
+        jobs[job_id]['done'] = True
+        processes.pop(job_id, None)
+
+@app.route('/stop/<job_id>', methods=['POST'])
+def stop_job(job_id):
+    proc = processes.get(job_id)
+    if proc and proc.poll() is None:
+        try:
+            proc.terminate()
+            try:
+                proc.wait(timeout=3)
+            except Exception:
+                proc.kill()
+        except Exception:
+            pass
+    jobs[job_id]['done'] = True
+    return jsonify({'stopped': True})
+
+@app.route('/', methods=['GET', 'POST'])
+def index():
+    if request.method == 'POST':
+        text = request.form.get('text', '')
+        # Handle uploaded text file
+        if 'text_file' in request.files and request.files['text_file']:
+            file = request.files['text_file']
+            if file.filename:
+                file_content = file.read().decode('utf-8')
+                text = file_content
+        if text:
+            filename = f"{uuid.uuid4()}.wav"
+            output_path = os.path.join(OUTPUT_FOLDER, filename)
+            cmd = [
+                'python', 'scripts/tortoise_tts.py',
+                '-o', output_path
+            ]
+            # Add text
+            if text:
+                cmd.append(text)
+            # Optional arguments
+            voice = request.form.get('voice')
+            if voice:
+                cmd.extend(['-v', voice])
+            preset = request.form.get('preset')
+            if preset:
+                cmd.extend(['-p', preset])
+            candidates = request.form.get('candidates')
+            if candidates and str(candidates) != '1':
+                cmd.extend(['--candidates', str(candidates)])
+            seed = request.form.get('seed')
+            if seed:
+                cmd.extend(['--seed', str(seed)])
+            device = request.form.get('device')
+            if device:
+                cmd.extend(['--device', device])
+            voices_dir = request.form.get('voices_dir')
+            if voices_dir:
+                cmd.extend(['-V', voices_dir])
+            text_split = request.form.get('text_split')
+            if text_split:
+                cmd.extend(['--text-split', text_split])
+            # Tuning options
+            if request.form.get('num_autoregressive_samples'):
+                cmd.extend(['--num-autoregressive-samples', str(request.form['num_autoregressive_samples'])])
+            if request.form.get('temperature'):
+                cmd.extend(['--temperature', str(request.form['temperature'])])
+            if request.form.get('length_penalty'):
+                cmd.extend(['--length-penalty', str(request.form['length_penalty'])])
+            if request.form.get('repetition_penalty'):
+                cmd.extend(['--repetition-penalty', str(request.form['repetition_penalty'])])
+            if request.form.get('top_p'):
+                cmd.extend(['--top-p', str(request.form['top_p'])])
+            if request.form.get('max_mel_tokens'):
+                cmd.extend(['--max-mel-tokens', str(request.form['max_mel_tokens'])])
+            if request.form.get('cvvp_amount'):
+                cmd.extend(['--cvvp-amount', str(request.form['cvvp_amount'])])
+            if request.form.get('diffusion_iterations'):
+                cmd.extend(['--diffusion-iterations', str(request.form['diffusion_iterations'])])
+            if request.form.get('cond_free'):
+                cmd.append('--cond-free')
+            if request.form.get('cond_free_k'):
+                cmd.extend(['--cond-free-k', str(request.form['cond_free_k'])])
+            if request.form.get('diffusion_temperature'):
+                cmd.extend(['--diffusion-temperature', str(request.form['diffusion_temperature'])])
+            if request.form.get('quiet'):
+                cmd.append('-q')
+            if request.form.get('produce_debug_state'):
+                cmd.append('--produce-debug-state')
+            if request.form.get('skip_existing'):
+                cmd.append('--skip-existing')
+            if request.form.get('disable_redaction'):
+                cmd.append('--disable-redaction')
+            env = os.environ.copy()
+            env["PYTHONPATH"] = os.path.abspath(os.path.dirname(__file__))
+            job_id = str(uuid.uuid4())
+            jobs[job_id] = {'progress': 0, 'log': '', 'done': False, 'audio_url': f"/audio/{filename}"}
+            t = threading.Thread(target=run_tts_with_progress, args=(cmd, env, job_id))
+            t.start()
+            return jsonify({'job_id': job_id})
+    # GET request
+    return render_template_string(HTML)
+
+@app.route('/progress/<job_id>')
+def progress(job_id):
+    job = jobs.get(job_id)
+    if not job:
+        return jsonify({'progress': 0, 'log': '', 'done': True, 'audio_url': None})
+    resp = {
+        'progress': job['progress'],
+        'log': job['log'],
+        'done': job['done'],
+        'audio_url': job['audio_url'] if job['done'] else None
+    }
+    return jsonify(resp)
+
+@app.route('/audio/<filename>')
+def audio(filename):
+    return send_file(os.path.join(OUTPUT_FOLDER, filename), as_attachment=False)
+
+if __name__ == '__main__':
+    app.run(debug=True, port=5000)
--- a/requirements-openui.txt
+++ b/requirements-openui.txt
@ -0,0 +1 @@
+Flask
--- a/requirements.txt
+++ b/requirements.txt
@ -18,7 +18,6 @@ appdirs
 nbconvert==5.3.1
 tornado==4.2
 pydantic==1.9.1
-deepspeed==0.8.3
 py-cpuinfo
 hjson
 psutil
--- a/scripts/tortoise_tts.py
+++ b/scripts/tortoise_tts.py
@ -8,9 +8,8 @@ import time

 import torch
 import torchaudio
-
 from tortoise.api import MODELS_DIR, TextToSpeech
-from tortoise.utils.audio import get_voices, load_voices, load_audio
+from tortoise.utils.audio import get_voices, load_audio, load_voices
 from tortoise.utils.text import split_and_recombine_text

 parser = argparse.ArgumentParser(
--- a/tortoise/api.py
+++ b/tortoise/api.py
@ -1,12 +1,9 @@
 import os
 import random
-import uuid
 from time import time
-from urllib import request

 import torch
 import torch.nn.functional as F
-import progressbar
 import torchaudio

 from tortoise.models.classifier import AudioMiniEncoderWithClassifierHead
--- a/tortoise/api_fast.py
+++ b/tortoise/api_fast.py
@ -1,29 +1,18 @@
 import os
 import random
-import uuid
 from time import time
-from urllib import request

 import torch
 import torch.nn.functional as F
-import progressbar
-import torchaudio
-import numpy as np
 from tortoise.models.classifier import AudioMiniEncoderWithClassifierHead
-from tortoise.models.diffusion_decoder import DiffusionTts
 from tortoise.models.autoregressive import UnifiedVoice
-from tqdm import tqdm
 from tortoise.models.arch_util import TorchMelSpectrogram
-from tortoise.models.clvp import CLVP
-from tortoise.models.cvvp import CVVP
 from tortoise.models.hifigan_decoder import HifiganGenerator
 from tortoise.models.random_latent_generator import RandomLatentConverter
-from tortoise.models.vocoder import UnivNetGenerator
-from tortoise.utils.audio import wav_to_univnet_mel, denormalize_tacotron_mel
+from tortoise.utils.audio import denormalize_tacotron_mel
 from tortoise.utils.diffusion import SpacedDiffusion, space_timesteps, get_named_beta_schedule
 from tortoise.utils.tokenizer import VoiceBpeTokenizer
 from tortoise.utils.wav2vec_alignment import Wav2VecAlignment
-from contextlib import contextmanager
 from tortoise.models.stream_generator import init_stream_support
 from huggingface_hub import hf_hub_download
 pbar = None
--- a/tortoise/models/transformer.py
+++ b/tortoise/models/transformer.py
@ -1,9 +1,7 @@
-from functools import partial

 import torch
 import torch.nn.functional as F
 from einops import rearrange
-from rotary_embedding_torch import RotaryEmbedding
 from torch import nn


--- a/tortoise/tts_stream.py
+++ b/tortoise/tts_stream.py
@ -3,7 +3,6 @@ import os
 from time import time

 import torch
-import torchaudio

 from api_fast import TextToSpeech, MODELS_DIR
 from utils.audio import load_audio, load_voices
--- a/tortoise/utils/wav2vec_alignment.py
+++ b/tortoise/utils/wav2vec_alignment.py
@ -1,10 +1,8 @@
-import re

 import torch
 import torchaudio
-from transformers import Wav2Vec2ForCTC, Wav2Vec2FeatureExtractor, Wav2Vec2CTCTokenizer, Wav2Vec2Processor
+from transformers import Wav2Vec2ForCTC, Wav2Vec2FeatureExtractor, Wav2Vec2CTCTokenizer

-from tortoise.utils.audio import load_audio


 def max_alignment(s1, s2, skip_character='~', record=None):