Upload 8 files

.gitignore

@@ -0,0 +1,7 @@
+.venv
+.aider*
+.env
+__pycache__
+.gradio
+output/*
+!output/.gitkeep

Dockerfile

@@ -0,0 +1,25 @@
+FROM pytorch/pytorch:2.1.0-cuda12.1-cudnn8-runtime
+
+ENV GRADIO_SERVER_PORT=7860
+ENV GRADIO_SERVER_NAME="0.0.0.0"
+
+# Create mount point and set permissions for persistent storage
+RUN mkdir -p /mnt && \
+    chown -R 1000:1000 /mnt && \
+    chmod 755 /mnt
+VOLUME /mnt
+
+WORKDIR /opt/app
+
+# Copy requirements first for better caching
+COPY requirements.txt requirements.txt
+
+RUN apt-get update && apt-get install -y git && \
+    pip install -r requirements.txt
+
+# Copy application files
+COPY app.py .
+COPY public/ /opt/app/public/
+
+# Run the Gradio app
+CMD ["python", "app.py"]

SDL.yaml

@@ -0,0 +1,60 @@
+---
+version: "2.0"
+services:
+  bark-small:
+    image: alexpedersen/audio-akash:0.1.3
+    expose:
+      - port: 7860
+        as: 7860
+        to:
+          - global: true
+        # accept:
+        #   - cars.ingress.europlots.com
+    params:
+      storage:
+        data:
+          mount: /mnt/
+          readOnly: false
+    # Optional but recommended
+    env:
+      - OUTPUT_DIR=/mnt/output # Use persistent storage for app generated files
+      - HF_HOME=/mnt/huggingface # Use persistent storage for model cache
+      # - PUBLIC_URL=cars.ingress.europlots.com
+profiles:
+  compute:
+    bark-small:
+      resources:
+        cpu:
+          units: 6
+        memory:
+          size: 16Gi
+        storage:
+          - size: 4GB
+          - name: data
+            size: 40GB
+            attributes:
+              persistent: true
+              class: beta3
+        gpu:
+          units: 1
+          attributes:
+            vendor:
+              nvidia:
+                # min 8gb GPU, for example:
+                # - model: rtx4090
+                # - model: rtx4080
+                # - model: rtx4070
+                # - model: rtx3090
+                # - model: rtx3080
+                # - model: rtx3070
+  placement:
+    dcloud:
+      pricing:
+        bark-small:
+          denom: uakt
+          amount: 1000
+deployment:
+  bark-small:
+    dcloud:
+      profile: bark-small
+      count: 1

app.py

@@ -0,0 +1,276 @@
+import os
+import time
+import glob
+from datetime import datetime, timedelta
+
+import numpy as np
+import torch
+from scipy.io.wavfile import write as write_wav
+
+from transformers import AutoProcessor, AutoModelForTextToWaveform, BarkModel
+
+import gradio as gr
+from fastapi import FastAPI
+from fastapi.staticfiles import StaticFiles
+import uvicorn
+from apscheduler.schedulers.background import BackgroundScheduler
+
+OUTPUT_DIR = os.environ.get("OUTPUT_DIR", "output")
+os.makedirs(OUTPUT_DIR, exist_ok=True)
+os.environ.update({
+    "SUNO_OFFLOAD_CPU": "True",
+    "SUNO_USE_SMALL_MODELS": "True"
+})
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+processor = AutoProcessor.from_pretrained("suno/bark-small")
+model = (BarkModel.from_pretrained("suno/bark-small", torch_dtype=torch.float16)
+         .to(device)
+         .to_bettertransformer())
+
+def create_bark_audio(text, voice_preset, device):
+    inputs = processor(text, voice_preset=voice_preset)
+    inputs = {k: v.to(device) if hasattr(v, 'to') else v for k, v in inputs.items()}
+    audio_array = model.generate(**inputs)
+    return audio_array.cpu().numpy().squeeze(), model.generation_config.sample_rate
+
+def save_audio(audio_array, sample_rate, prefix="audio"):
+    audio_array = np.clip(audio_array.astype(np.float32), -1, 1)
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    filename = os.path.join(OUTPUT_DIR, f"{prefix}_{timestamp}.wav")
+    write_wav(filename, sample_rate, audio_array)
+    return filename
+
+def generate_speech(text, voice_preset="v2/en_speaker_6"):
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    audio_array, sample_rate = create_bark_audio(text, voice_preset, device)
+    return save_audio(audio_array, sample_rate)
+
+def text_to_speech_with_url(text, voice):
+    audio_file = generate_speech(text, VOICES[voice])
+    filename = os.path.basename(audio_file)
+    base_url = os.environ.get("PUBLIC_URL", "http://localhost:7860")
+    return audio_file, f"{base_url}/generated/{filename}"
+
+def cleanup_old_files():
+    cutoff_time = datetime.now() - timedelta(hours=24)
+    for file in glob.glob(os.path.join(OUTPUT_DIR, "audio_*.wav")):
+        if datetime.fromtimestamp(os.path.getmtime(file)) < cutoff_time:
+            try:
+                os.remove(file)
+            except Exception as e:
+                print(f"Error removing file {file}: {e}")
+
+VOICES = {
+    'Speaker 0 (EN)':'v2/en_speaker_0',
+    'Speaker 1 (EN)':'v2/en_speaker_1',
+    'Speaker 2 (EN)':'v2/en_speaker_2',
+    'Speaker 3 (EN)':'v2/en_speaker_3',
+    'Speaker 4 (EN)':'v2/en_speaker_4',
+    'Speaker 5 (EN)':'v2/en_speaker_5',
+    'Speaker 6 (EN)':'v2/en_speaker_6',
+    'Speaker 7 (EN)':'v2/en_speaker_7',
+    'Speaker 8 (EN)':'v2/en_speaker_8',
+    'Speaker 9 (EN)':'v2/en_speaker_9',
+    'Speaker 0 (ZH)':'v2/zh_speaker_0',
+    'Speaker 1 (ZH)':'v2/zh_speaker_1',
+    'Speaker 2 (ZH)':'v2/zh_speaker_2',
+    'Speaker 3 (ZH)':'v2/zh_speaker_3',
+    'Speaker 4 (ZH)':'v2/zh_speaker_4',
+    'Speaker 5 (ZH)':'v2/zh_speaker_5',
+    'Speaker 6 (ZH)':'v2/zh_speaker_6',
+    'Speaker 7 (ZH)':'v2/zh_speaker_7',
+    'Speaker 8 (ZH)':'v2/zh_speaker_8',
+    'Speaker 9 (ZH)':'v2/zh_speaker_9',
+    'Speaker 0 (FR)':'v2/fr_speaker_0',
+    'Speaker 1 (FR)':'v2/fr_speaker_1',
+    'Speaker 2 (FR)':'v2/fr_speaker_2',
+    'Speaker 3 (FR)':'v2/fr_speaker_3',
+    'Speaker 4 (FR)':'v2/fr_speaker_4',
+    'Speaker 5 (FR)':'v2/fr_speaker_5',
+    'Speaker 6 (FR)':'v2/fr_speaker_6',
+    'Speaker 7 (FR)':'v2/fr_speaker_7',
+    'Speaker 8 (FR)':'v2/fr_speaker_8',
+    'Speaker 9 (FR)':'v2/fr_speaker_9',
+    'Speaker 0 (DE)':'v2/de_speaker_0',
+    'Speaker 1 (DE)':'v2/de_speaker_1',
+    'Speaker 2 (DE)':'v2/de_speaker_2',
+    'Speaker 3 (DE)':'v2/de_speaker_3',
+    'Speaker 4 (DE)':'v2/de_speaker_4',
+    'Speaker 5 (DE)':'v2/de_speaker_5',
+    'Speaker 6 (DE)':'v2/de_speaker_6',
+    'Speaker 7 (DE)':'v2/de_speaker_7',
+    'Speaker 8 (DE)':'v2/de_speaker_8',
+    'Speaker 9 (DE)':'v2/de_speaker_9',
+    'Speaker 0 (HI)':'v2/hi_speaker_0',
+    'Speaker 1 (HI)':'v2/hi_speaker_1',
+    'Speaker 2 (HI)':'v2/hi_speaker_2',
+    'Speaker 3 (HI)':'v2/hi_speaker_3',
+    'Speaker 4 (HI)':'v2/hi_speaker_4',
+    'Speaker 5 (HI)':'v2/hi_speaker_5',
+    'Speaker 6 (HI)':'v2/hi_speaker_6',
+    'Speaker 7 (HI)':'v2/hi_speaker_7',
+    'Speaker 8 (HI)':'v2/hi_speaker_8',
+    'Speaker 9 (HI)':'v2/hi_speaker_9',
+    'Speaker 0 (IT)':'v2/it_speaker_0',
+    'Speaker 1 (IT)':'v2/it_speaker_1',
+    'Speaker 2 (IT)':'v2/it_speaker_2',
+    'Speaker 3 (IT)':'v2/it_speaker_3',
+    'Speaker 4 (IT)':'v2/it_speaker_4',
+    'Speaker 5 (IT)':'v2/it_speaker_5',
+    'Speaker 6 (IT)':'v2/it_speaker_6',
+    'Speaker 7 (IT)':'v2/it_speaker_7',
+    'Speaker 8 (IT)':'v2/it_speaker_8',
+    'Speaker 9 (IT)':'v2/it_speaker_9',
+    'Speaker 0 (JA)':'v2/ja_speaker_0',
+    'Speaker 1 (JA)':'v2/ja_speaker_1',
+    'Speaker 2 (JA)':'v2/ja_speaker_2',
+    'Speaker 3 (JA)':'v2/ja_speaker_3',
+    'Speaker 4 (JA)':'v2/ja_speaker_4',
+    'Speaker 5 (JA)':'v2/ja_speaker_5',
+    'Speaker 6 (JA)':'v2/ja_speaker_6',
+    'Speaker 7 (JA)':'v2/ja_speaker_7',
+    'Speaker 8 (JA)':'v2/ja_speaker_8',
+    'Speaker 9 (JA)':'v2/ja_speaker_9',
+    'Speaker 0 (KO)':'v2/ko_speaker_0',
+    'Speaker 1 (KO)':'v2/ko_speaker_1',
+    'Speaker 2 (KO)':'v2/ko_speaker_2',
+    'Speaker 3 (KO)':'v2/ko_speaker_3',
+    'Speaker 4 (KO)':'v2/ko_speaker_4',
+    'Speaker 5 (KO)':'v2/ko_speaker_5',
+    'Speaker 6 (KO)':'v2/ko_speaker_6',
+    'Speaker 7 (KO)':'v2/ko_speaker_7',
+    'Speaker 8 (KO)':'v2/ko_speaker_8',
+    'Speaker 9 (KO)':'v2/ko_speaker_9',
+    'Speaker 0 (PL)':'v2/pl_speaker_0',
+    'Speaker 1 (PL)':'v2/pl_speaker_1',
+    'Speaker 2 (PL)':'v2/pl_speaker_2',
+    'Speaker 3 (PL)':'v2/pl_speaker_3',
+    'Speaker 4 (PL)':'v2/pl_speaker_4',
+    'Speaker 5 (PL)':'v2/pl_speaker_5',
+    'Speaker 6 (PL)':'v2/pl_speaker_6',
+    'Speaker 7 (PL)':'v2/pl_speaker_7',
+    'Speaker 8 (PL)':'v2/pl_speaker_8',
+    'Speaker 9 (PL)':'v2/pl_speaker_9',
+    'Speaker 0 (PT)':'v2/pt_speaker_0',
+    'Speaker 1 (PT)':'v2/pt_speaker_1',
+    'Speaker 2 (PT)':'v2/pt_speaker_2',
+    'Speaker 3 (PT)':'v2/pt_speaker_3',
+    'Speaker 4 (PT)':'v2/pt_speaker_4',
+    'Speaker 5 (PT)':'v2/pt_speaker_5',
+    'Speaker 6 (PT)':'v2/pt_speaker_6',
+    'Speaker 7 (PT)':'v2/pt_speaker_7',
+    'Speaker 8 (PT)':'v2/pt_speaker_8',
+    'Speaker 9 (PT)':'v2/pt_speaker_9',
+    'Speaker 0 (RU)':'v2/ru_speaker_0',
+    'Speaker 1 (RU)':'v2/ru_speaker_1',
+    'Speaker 2 (RU)':'v2/ru_speaker_2',
+    'Speaker 3 (RU)':'v2/ru_speaker_3',
+    'Speaker 4 (RU)':'v2/ru_speaker_4',
+    'Speaker 5 (RU)':'v2/ru_speaker_5',
+    'Speaker 6 (RU)':'v2/ru_speaker_6',
+    'Speaker 7 (RU)':'v2/ru_speaker_7',
+    'Speaker 8 (RU)':'v2/ru_speaker_8',
+    'Speaker 9 (RU)':'v2/ru_speaker_9',
+    'Speaker 0 (ES)':'v2/es_speaker_0',
+    'Speaker 1 (ES)':'v2/es_speaker_1',
+    'Speaker 2 (ES)':'v2/es_speaker_2',
+    'Speaker 3 (ES)':'v2/es_speaker_3',
+    'Speaker 4 (ES)':'v2/es_speaker_4',
+    'Speaker 5 (ES)':'v2/es_speaker_5',
+    'Speaker 6 (ES)':'v2/es_speaker_6',
+    'Speaker 7 (ES)':'v2/es_speaker_7',
+    'Speaker 8 (ES)':'v2/es_speaker_8',
+    'Speaker 9 (ES)':'v2/es_speaker_9',
+    'Speaker 0 (TR)':'v2/tr_speaker_0',
+    'Speaker 1 (TR)':'v2/tr_speaker_1',
+    'Speaker 2 (TR)':'v2/tr_speaker_2',
+    'Speaker 3 (TR)':'v2/tr_speaker_3',
+    'Speaker 4 (TR)':'v2/tr_speaker_4',
+    'Speaker 5 (TR)':'v2/tr_speaker_5',
+    'Speaker 6 (TR)':'v2/tr_speaker_6',
+    'Speaker 7 (TR)':'v2/tr_speaker_7',
+    'Speaker 8 (TR)':'v2/tr_speaker_8',
+    'Speaker 9 (TR)':'v2/tr_speaker_9',
+}
+
+CUSTOM_CSS = """
+#component-16 { display: none !important; }
+.gradio-container .main h1 { padding-top: 60px; position: relative; }
+.gradio-container .main h1::before {
+    content: '';
+    position: absolute;
+    top: 0;
+    left: 50%;
+    transform: translateX(-50%);
+    width: 253px;
+    height: 50px;
+    background-image: url('public/AkashLogo.svg');
+    background-repeat: no-repeat;
+    background-position: center;
+    background-size: contain;
+}
+"""
+
+with gr.Blocks(css=CUSTOM_CSS) as gradio_audio:
+    gr.Interface(
+        fn=text_to_speech_with_url,
+        inputs=[
+            gr.Textbox(label="Text to audio", placeholder="Enter text here...", show_copy_button=False),
+            gr.Dropdown(choices=list(VOICES.keys()), value="Speaker 0 (EN)", label="Voice")
+        ],
+        outputs=[
+            gr.Audio(label="Generated Speech"),
+            gr.Textbox(label="Public URL", interactive=False, show_copy_button=True)
+        ],
+        title="Audio Generator",
+        description="""
+        Transform text into natural-sounding speech using the Bark AI model. 
+        Features support for multiple languages and voice styles.
+        
+        **How to use:**
+        1. Enter your text in any supported language
+        2. Select a voice preset
+        3. Click submit to generate speech
+        4. Get the public URL to share/download the generated audio (it will expire in 24 hours)
+        """,
+        article="""<div style="text-align: center">Powered by <a href="https://huggingface.co/suno/bark-small">Bark-small</a> model and <a href="https://akash.network">Akash Network</a>, created by <a href="https://github.com/alexx855">alexx855</a></div>""",
+        examples=[
+            ["Welcome to the news. Today's top story...", "Speaker 0 (EN)"],
+            ["The quick brown fox jumps over the lazy dog.", "Speaker 1 (EN)"],
+            ["你好，今天天气真不错。", "Speaker 0 (ZH)"],
+            ["Bonjour, comment allez-vous aujourd'hui?", "Speaker 0 (FR)"],
+            ["J'aime beaucoup voyager en France.", "Speaker 1 (FR)"],
+            ["Guten Tag, wie geht es Ihnen?", "Speaker 0 (DE)"],
+            ["Das Wetter ist heute sehr schön.", "Speaker 1 (DE)"],
+            ["नमस्ते, आप कैसे हैं?", "Speaker 0 (HI)"],
+            ["मौसम बहुत सुहावन�� है।", "Speaker 1 (HI)"],
+            ["Buongiorno, come stai oggi?", "Speaker 0 (IT)"],
+            ["Mi piace molto viaggiare in Italia.", "Speaker 1 (IT)"],
+            ["こんにちは、お元気ですか？", "Speaker 0 (JA)"],
+            ["今日はとても良い天気ですね。", "Speaker 1 (JA)"],
+            ["안녕하세요, 오늘 기분이 어떠신가요?", "Speaker 0 (KO)"],
+            ["날씨가 정말 좋네요.", "Speaker 1 (KO)"],
+            ["Dzień dobry, jak się masz?", "Speaker 0 (PL)"],
+            ["Dzisiaj jest bardzo ładna pogoda.", "Speaker 1 (PL)"],
+            ["Olá, como está você hoje?", "Speaker 0 (PT)"],
+            ["O tempo está muito bonito hoje.", "Speaker 1 (PT)"],
+            ["Здравствуйте, как ваши дела?", "Speaker 0 (RU)"],
+            ["Сегодня прекрасная погода.", "Speaker 1 (RU)"],
+            ["Hola, ¿cómo estás hoy?", "Speaker 0 (ES)"],
+            ["El tiempo está muy bonito hoy.", "Speaker 1 (ES)"],
+            ["Merhaba, bugün nasılsınız?", "Speaker 0 (TR)"],
+            ["Bugün hava çok güzel.", "Speaker 1 (TR)"]
+        ]
+    )
+
+scheduler = BackgroundScheduler()
+scheduler.add_job(cleanup_old_files, 'interval', hours=1)
+scheduler.start()
+
+if __name__ == "__main__":
+    app = FastAPI()
+    app.mount("/generated", StaticFiles(directory=OUTPUT_DIR), name="generated")
+    app.mount("/public", StaticFiles(directory="public"), name="public")
+    gradio_app = gr.mount_gradio_app(app, gradio_audio, path="/", favicon_path="public/favicon.ico")
+    uvicorn.run(app, host="0.0.0.0", port=7860)

main.py

@@ -0,0 +1,115 @@
+import torch
+from transformers import AutoProcessor, AutoModelForTextToWaveform, BarkModel
+from scipy.io.wavfile import write as write_wav
+import os
+import time
+from datetime import datetime, timedelta
+import numpy as np
+from apscheduler.schedulers.background import BackgroundScheduler
+import glob
+
+# Environment settings
+os.environ["SUNO_OFFLOAD_CPU"] = "True"
+os.environ["SUNO_USE_SMALL_MODELS"] = "True"
+
+# Create output directory if it doesn't exist
+OUTPUT_DIR = os.environ.get("OUTPUT_DIR", "output") 
+os.makedirs(OUTPUT_DIR, exist_ok=True)
+
+#create hf directory if it doesn't exist
+HF_DIR = os.environ.get("HF_HOME", "~/.cache/huggingface")
+
+def log_time(start_time, step_name):
+    elapsed = time.time() - start_time
+    print(f"{step_name}: {elapsed:.2f} seconds")
+    return time.time()
+
+start = time.time()
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+processor = AutoProcessor.from_pretrained("suno/bark-small")
+model = BarkModel.from_pretrained("suno/bark-small", torch_dtype=torch.float16).to(device)
+model = model.to_bettertransformer()
+model.enable_cpu_offload()
+
+start = log_time(start, "Model loading")
+
+# download and load all models
+# preload_models()
+
+def cleanup_old_files():
+    """Remove audio files older than 24 hour"""
+    cutoff_time = datetime.now() - timedelta(hours=24) 
+    for file in glob.glob(os.path.join(OUTPUT_DIR, "audio_*.wav")):
+        file_time = datetime.fromtimestamp(os.path.getmtime(file))
+        if file_time < cutoff_time:
+            try:
+                os.remove(file)
+                print(f"Removed old file: {file}")
+            except Exception as e:
+                print(f"Error removing file {file}: {e}")
+
+# Initialize scheduler
+scheduler = BackgroundScheduler()
+scheduler.add_job(cleanup_old_files, 'interval', hours=1) 
+scheduler.start()
+
+
+def create_bark_audio(text, voice_preset, device):
+    try:
+        start = time.time()
+        # Process input text directly without reloading model
+        inputs = processor(
+            text,
+            voice_preset=voice_preset,
+        )
+        # Move inputs to device
+        inputs = {k: v.to(device) if hasattr(v, 'to') else v for k, v in inputs.items()}
+        start = log_time(start, "Input processing")
+
+        # Generate audio
+        start = time.time()
+        audio_array = model.generate(**inputs)
+        audio_array = audio_array.cpu().numpy().squeeze()
+
+        start = log_time(start, "Audio generation")
+
+        return audio_array, model.generation_config.sample_rate
+    
+    except Exception as e:
+        print(f"Error during audio generation: {str(e)}")
+        raise
+
+def save_audio(audio_array, sample_rate, prefix="audio"):
+    try:
+        start = time.time()
+        # Convert to float32 and normalize
+        audio_array = audio_array.astype(np.float32)
+        # Ensure audio is in the range [-1, 1]
+        audio_array = np.clip(audio_array, -1, 1)
+        
+        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        filename = os.path.join(OUTPUT_DIR, f"{prefix}_{timestamp}.wav")
+        write_wav(filename, sample_rate, audio_array)
+        log_time(start, "Audio saving")
+        return filename
+    
+    except Exception as e:
+        print(f"Error saving audio file: {str(e)}")
+        raise
+
+def generate_speech(text, voice_preset="v2/en_speaker_6"):
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+
+    try:
+        audio_array, sample_rate = create_bark_audio(text, voice_preset, device)
+        filename = save_audio(audio_array, sample_rate)
+        return filename
+    except Exception as e:
+        print(f"An error occurred: {str(e)}")
+        raise
+
+if __name__ == "__main__":
+    text = "my cat is very cute"
+    filename = generate_speech(text)
+    print(f"Audio saved as: {filename}")

requirements.txt

@@ -0,0 +1,9 @@
+git+https://github.com/huggingface/transformers.git
+git+https://github.com/huggingface/accelerate
+git+https://github.com/huggingface/optimum.git
+git+https://github.com/suno-ai/bark.git
+torch
+scipy
+numpy
+gradio
+apscheduler