import os
import re
import torch
import tempfile
import logging
import math
from typing import Tuple, Union, Any
from scipy.io.wavfile import write
from pydub import AudioSegment
from dotenv import load_dotenv
import spaces
import gradio as gr
import numpy as np
# Transformers & Models
from transformers import (
AutoTokenizer,
AutoModelForCausalLM,
pipeline,
AutoProcessor,
MusicgenForConditionalGeneration,
)
# Coqui TTS
from TTS.api import TTS
# Diffusers for sound design generation
from diffusers import DiffusionPipeline, AudioLDMPipeline
import diffusers
from packaging import version
# ---------------------------------------------------------------------
# Setup Logging and Environment Variables
# ---------------------------------------------------------------------
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
load_dotenv()
HF_TOKEN = os.getenv("HF_TOKEN")
if not HF_TOKEN:
logging.warning("HF_TOKEN is not set in your environment. Some model downloads might fail.")
# ---------------------------------------------------------------------
# Global Model Caches
# ---------------------------------------------------------------------
LLAMA_PIPELINES: dict[str, Any] = {}
MUSICGEN_MODELS: dict[str, Any] = {}
TTS_MODELS: dict[str, Any] = {}
SOUND_DESIGN_PIPELINES: dict[str, Any] = {}
# ---------------------------------------------------------------------
# Utility Functions
# ---------------------------------------------------------------------
def clean_text(text: str) -> str:
"""
Remove undesired characters that may not be recognized by the model.
Args:
text (str): Input text to be cleaned.
Returns:
str: Cleaned text.
"""
return re.sub(r'\*', '', text)
# ---------------------------------------------------------------------
# Model Helper Functions
# ---------------------------------------------------------------------
def get_llama_pipeline(model_id: str, token: str) -> Any:
"""
Returns a cached LLaMA text-generation pipeline or loads a new one.
Args:
model_id (str): Hugging Face model ID.
token (str): Hugging Face token.
Returns:
Any: A Hugging Face text-generation pipeline.
"""
if model_id in LLAMA_PIPELINES:
return LLAMA_PIPELINES[model_id]
logging.info(f"Loading LLaMA model from {model_id}...")
tokenizer = AutoTokenizer.from_pretrained(model_id, use_auth_token=token)
model = AutoModelForCausalLM.from_pretrained(
model_id,
use_auth_token=token,
torch_dtype=torch.float16,
device_map="auto",
trust_remote_code=True,
)
text_pipeline = pipeline("text-generation", model=model, tokenizer=tokenizer)
LLAMA_PIPELINES[model_id] = text_pipeline
return text_pipeline
def get_musicgen_model(model_key: str = "facebook/musicgen-large") -> Tuple[Any, Any]:
"""
Returns a cached MusicGen model and processor, or loads new ones.
Args:
model_key (str): Hugging Face model key (default is 'facebook/musicgen-large').
Returns:
Tuple[Any, Any]: The MusicGen model and its processor.
"""
if model_key in MUSICGEN_MODELS:
return MUSICGEN_MODELS[model_key]
logging.info(f"Loading MusicGen model from {model_key}...")
model = MusicgenForConditionalGeneration.from_pretrained(model_key)
processor = AutoProcessor.from_pretrained(model_key)
device = "cuda" if torch.cuda.is_available() else "cpu"
model.to(device)
MUSICGEN_MODELS[model_key] = (model, processor)
return model, processor
def get_tts_model(model_name: str = "tts_models/en/ljspeech/tacotron2-DDC") -> TTS:
"""
Returns a cached TTS model or loads a new one.
Args:
model_name (str): Identifier for the TTS model.
Returns:
TTS: A Coqui TTS model.
"""
if model_name in TTS_MODELS:
return TTS_MODELS[model_name]
logging.info(f"Loading TTS model: {model_name}...")
tts_model = TTS(model_name)
TTS_MODELS[model_name] = tts_model
return tts_model
def get_sound_design_pipeline(model_name: str, token: str) -> Any:
"""
Returns a cached DiffusionPipeline for sound design, or loads a new one.
Raises an error if diffusers version is less than 0.21.0.
Args:
model_name (str): The model name to load.
token (str): Hugging Face token.
Returns:
Any: A DiffusionPipeline for sound design.
Raises:
ValueError: If diffusers version is lower than 0.21.0.
"""
if version.parse(diffusers.__version__) < version.parse("0.21.0"):
raise ValueError("AudioLDM2 requires diffusers>=0.21.0. Please upgrade your diffusers package.")
if model_name in SOUND_DESIGN_PIPELINES:
return SOUND_DESIGN_PIPELINES[model_name]
logging.info(f"Loading sound design pipeline from {model_name}...")
pipe = DiffusionPipeline.from_pretrained(
model_name,
pipeline_class=AudioLDMPipeline,
use_auth_token=token
)
SOUND_DESIGN_PIPELINES[model_name] = pipe
return pipe
# ---------------------------------------------------------------------
# Script Generation Function
# ---------------------------------------------------------------------
@spaces.GPU(duration=100)
def generate_script(user_prompt: str, model_id: str, token: str, duration: int) -> Tuple[str, str, str]:
"""
Generates a voice-over script, sound design suggestions, and music ideas based on the user prompt.
Args:
user_prompt (str): The user-provided concept.
model_id (str): The LLaMA model ID.
token (str): Hugging Face token.
duration (int): The desired duration in seconds.
Returns:
Tuple[str, str, str]: Voice-over script, sound design suggestions, and music suggestions.
"""
try:
text_pipeline = get_llama_pipeline(model_id, token)
system_prompt = (
"You are an expert radio imaging producer specializing in sound design and music. "
f"Based on the user's concept and the selected duration of {duration} seconds, produce the following:\n"
"1. A concise voice-over script. Prefix this section with 'Voice-Over Script:'\n"
"2. Suggestions for sound design. Prefix this section with 'Sound Design Suggestions:'\n"
"3. Music styles or track recommendations. Prefix this section with 'Music Suggestions:'"
)
combined_prompt = f"{system_prompt}\nUser concept: {user_prompt}\nOutput:"
with torch.inference_mode():
result = text_pipeline(
combined_prompt,
max_new_tokens=300,
do_sample=True,
temperature=0.8
)
generated_text = result[0]["generated_text"]
if "Output:" in generated_text:
generated_text = generated_text.split("Output:")[-1].strip()
# Extract sections using regex
pattern = r"Voice-Over Script:\s*(.*?)\s*Sound Design Suggestions:\s*(.*?)\s*Music Suggestions:\s*(.*)"
match = re.search(pattern, generated_text, re.DOTALL)
if match:
voice_script, sound_design, music_suggestions = (grp.strip() for grp in match.groups())
else:
voice_script = "No voice-over script found."
sound_design = "No sound design suggestions found."
music_suggestions = "No music suggestions found."
return voice_script, sound_design, music_suggestions
except Exception as e:
logging.exception("Error generating script")
return f"Error generating script: {e}", "", ""
# ---------------------------------------------------------------------
# Voice-Over Generation Function
# ---------------------------------------------------------------------
@spaces.GPU(duration=100)
def generate_voice(script: str, tts_model_name: str = "tts_models/en/ljspeech/tacotron2-DDC") -> Union[str, Any]:
"""
Generates a voice-over audio file from a script using Coqui TTS.
Args:
script (str): The voice-over script.
tts_model_name (str): The TTS model name.
Returns:
Union[str, Any]: The file path to the generated .wav file or an error message.
"""
try:
if not script.strip():
return "Error: No script provided."
cleaned_script = clean_text(script)
tts_model = get_tts_model(tts_model_name)
output_path = os.path.join(tempfile.gettempdir(), "voice_over.wav")
tts_model.tts_to_file(text=cleaned_script, file_path=output_path)
return output_path
except Exception as e:
logging.exception("Error generating voice")
return f"Error generating voice: {e}"
# ---------------------------------------------------------------------
# Music Generation Function
# ---------------------------------------------------------------------
@spaces.GPU(duration=200)
def generate_music(prompt: str, audio_length: int) -> Union[str, Any]:
"""
Generates a music track using the MusicGen model based on the prompt.
Args:
prompt (str): Music suggestion prompt.
audio_length (int): Number of tokens determining audio length.
Returns:
Union[str, Any]: The file path to the generated .wav file or an error message.
"""
try:
if not prompt.strip():
return "Error: No music suggestion provided."
model_key = "facebook/musicgen-large"
musicgen_model, musicgen_processor = get_musicgen_model(model_key)
device = "cuda" if torch.cuda.is_available() else "cpu"
inputs = musicgen_processor(text=[prompt], padding=True, return_tensors="pt").to(device)
with torch.inference_mode():
outputs = musicgen_model.generate(**inputs, max_new_tokens=audio_length)
audio_data = outputs[0, 0].cpu().numpy()
# Normalize audio data to 16-bit integer range
normalized_audio = (audio_data / np.max(np.abs(audio_data)) * 32767).astype("int16")
output_path = os.path.join(tempfile.gettempdir(), "musicgen_large_generated_music.wav")
write(output_path, 44100, normalized_audio)
return output_path
except Exception as e:
logging.exception("Error generating music")
return f"Error generating music: {e}"
# ---------------------------------------------------------------------
# Sound Design Generation Function
# ---------------------------------------------------------------------
@spaces.GPU(duration=200)
def generate_sound_design(prompt: str) -> Union[str, Any]:
"""
Generates a sound design audio file using AudioLDM 2 based on the prompt.
Args:
prompt (str): Sound design prompt.
Returns:
Union[str, Any]: The file path to the generated .wav file or an error message.
"""
try:
if not prompt.strip():
return "Error: No sound design suggestion provided."
pipe = get_sound_design_pipeline("cvssp/audioldm2", HF_TOKEN)
result = pipe(prompt) # Expected to return a dict with key 'audios'
audio_samples = result["audios"][0]
normalized_audio = (audio_samples / np.max(np.abs(audio_samples)) * 32767).astype("int16")
output_path = os.path.join(tempfile.gettempdir(), "sound_design_generated.wav")
write(output_path, 44100, normalized_audio)
return output_path
except Exception as e:
logging.exception("Error generating sound design")
return f"Error generating sound design: {e}"
# ---------------------------------------------------------------------
# Audio Blending Function
# ---------------------------------------------------------------------
@spaces.GPU(duration=100)
def blend_audio(voice_path: str, sound_effect_path: str, music_path: str, ducking: bool, duck_level: int = 10) -> Union[str, Any]:
"""
Blends three audio files (voice, sound design, and music) by:
- Looping/trimming music and sound design to match voice duration.
- Optionally applying ducking to background tracks.
- Overlaying the voice on top of the background.
Args:
voice_path (str): Path to the voice audio file.
sound_effect_path (str): Path to the sound design audio file.
music_path (str): Path to the music audio file.
ducking (bool): Whether to apply ducking.
duck_level (int): Amount of attenuation in dB.
Returns:
Union[str, Any]: The file path to the blended .wav file or an error message.
"""
try:
for path in [voice_path, sound_effect_path, music_path]:
if not os.path.isfile(path):
return f"Error: Missing audio file for {path}"
# Load audio segments
voice = AudioSegment.from_wav(voice_path)
music = AudioSegment.from_wav(music_path)
sound_effect = AudioSegment.from_wav(sound_effect_path)
voice_len = len(voice) # duration in milliseconds
# Loop or trim music to match voice duration using pydub multiplication
if len(music) < voice_len:
repeats = math.ceil(voice_len / len(music))
music = (music * repeats)[:voice_len]
else:
music = music[:voice_len]
# Loop or trim sound design to match voice duration
if len(sound_effect) < voice_len:
repeats = math.ceil(voice_len / len(sound_effect))
sound_effect = (sound_effect * repeats)[:voice_len]
else:
sound_effect = sound_effect[:voice_len]
# Apply ducking if enabled
if ducking:
music = music - duck_level
sound_effect = sound_effect - duck_level
# Overlay music and sound effect for background
background = music.overlay(sound_effect)
# Overlay voice on top of background
final_audio = background.overlay(voice)
output_path = os.path.join(tempfile.gettempdir(), "blended_output.wav")
final_audio.export(output_path, format="wav")
return output_path
except Exception as e:
logging.exception("Error blending audio")
return f"Error blending audio: {e}"
# ---------------------------------------------------------------------
# Gradio Interface
# ---------------------------------------------------------------------
with gr.Blocks(css="""
/* Global Styles */
body {
background: linear-gradient(135deg, #1d1f21, #3a3d41);
color: #f0f0f0;
font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;
}
.header {
text-align: center;
padding: 2rem 1rem;
background: linear-gradient(90deg, #6a11cb, #2575fc);
border-radius: 0 0 20px 20px;
margin-bottom: 2rem;
}
.header h1 {
margin: 0;
font-size: 2.5rem;
}
.header p {
font-size: 1.2rem;
}
.gradio-container {
background: #2e2e2e;
border-radius: 10px;
padding: 1rem;
}
.tab-title {
font-size: 1.1rem;
font-weight: bold;
}
.footer {
text-align: center;
font-size: 0.9em;
margin-top: 2rem;
padding: 1rem;
color: #cccccc;
}
""") as demo:
# Custom Header
with gr.Row(elem_classes="header"):
gr.Markdown("""
<h1>🎧 Ai Ads Promo</h1>
<p>Your all-in-one AI solution for creating professional audio ads.</p>
""")
gr.Markdown("""
**Welcome to Ai Ads Promo!**
This app helps you create amazing audio ads in just a few steps:
1. **Script Generation:** Provide your idea and get a voice-over script, sound design, and music suggestions.
2. **Voice Synthesis:** Convert the script into natural-sounding speech.
3. **Music Production:** Generate a custom music track.
4. **Sound Design:** Create creative sound effects.
5. **Audio Blending:** Seamlessly blend voice, music, and sound design (with optional ducking).
""")
with gr.Tabs():
# Step 1: Script Generation
with gr.Tab("📝 Script Generation"):
with gr.Row():
user_prompt = gr.Textbox(
label="Promo Ads Idea",
placeholder="E.g., A 30-second ad for a radio morning show...",
lines=2
)
with gr.Row():
llama_model_id = gr.Textbox(
label="LLaMA Model ID",
value="meta-llama/Meta-Llama-3-8B-Instruct",
placeholder="Enter a valid Hugging Face model ID"
)
duration = gr.Slider(
label="Desired Ad Duration (seconds)",
minimum=15,
maximum=60,
step=15,
value=30
)
generate_script_button = gr.Button("Generate Script", variant="primary")
script_output = gr.Textbox(label="Generated Voice-Over Script", lines=5, interactive=False)
sound_design_output = gr.Textbox(label="Sound Design Suggestions", lines=3, interactive=False)
music_suggestion_output = gr.Textbox(label="Music Suggestions", lines=3, interactive=False)
generate_script_button.click(
fn=lambda prompt, model_id, dur: generate_script(prompt, model_id, HF_TOKEN, dur),
inputs=[user_prompt, llama_model_id, duration],
outputs=[script_output, sound_design_output, music_suggestion_output],
)
# Step 2: Voice Synthesis
with gr.Tab("🎤 Voice Synthesis"):
gr.Markdown("Generate a natural-sounding voice-over using Coqui TTS.")
selected_tts_model = gr.Dropdown(
label="TTS Model",
choices=[
"tts_models/en/ljspeech/tacotron2-DDC",
"tts_models/en/ljspeech/vits",
"tts_models/en/sam/tacotron-DDC",
],
value="tts_models/en/ljspeech/tacotron2-DDC",
multiselect=False
)
generate_voice_button = gr.Button("Generate Voice-Over", variant="primary")
voice_audio_output = gr.Audio(label="Voice-Over (WAV)", type="filepath")
generate_voice_button.click(
fn=lambda script, tts_model: generate_voice(script, tts_model),
inputs=[script_output, selected_tts_model],
outputs=voice_audio_output,
)
# Step 3: Music Production
with gr.Tab("🎶 Music Production"):
gr.Markdown("Generate a custom music track using the **MusicGen Large** model.")
audio_length = gr.Slider(
label="Music Length (tokens)",
minimum=128,
maximum=1024,
step=64,
value=512,
info="Increase tokens for longer audio (inference time may vary)."
)
generate_music_button = gr.Button("Generate Music", variant="primary")
music_output = gr.Audio(label="Generated Music (WAV)", type="filepath")
generate_music_button.click(
fn=lambda music_prompt, length: generate_music(music_prompt, length),
inputs=[music_suggestion_output, audio_length],
outputs=[music_output],
)
# Step 4: Sound Design Generation
with gr.Tab("🎧 Sound Design Generation"):
gr.Markdown("Generate a creative sound design track based on the script's suggestions.")
generate_sound_design_button = gr.Button("Generate Sound Design", variant="primary")
sound_design_audio_output = gr.Audio(label="Generated Sound Design (WAV)", type="filepath")
generate_sound_design_button.click(
fn=generate_sound_design,
inputs=[sound_design_output],
outputs=[sound_design_audio_output],
)
# Step 5: Audio Blending (Voice + Sound Design + Music)
with gr.Tab("🎚️ Audio Blending"):
gr.Markdown("Blend your voice-over, sound design, and music track. Enable ducking to lower background audio during voice segments.")
ducking_checkbox = gr.Checkbox(label="Enable Ducking?", value=True)
duck_level_slider = gr.Slider(
label="Ducking Level (dB attenuation)",
minimum=0,
maximum=20,
step=1,
value=10
)
blend_button = gr.Button("Blend Audio", variant="primary")
blended_output = gr.Audio(label="Final Blended Output (WAV)", type="filepath")
blend_button.click(
fn=blend_audio,
inputs=[voice_audio_output, sound_design_audio_output, music_output, ducking_checkbox, duck_level_slider],
outputs=blended_output
)
# Footer and Visitor Badge
gr.Markdown("""
<div class="footer">
<hr>
Created with ❤️ by <a href="https://bilsimaging.com" target="_blank" style="color: #88aaff;">bilsimaging.com</a>
<br>
<small>Ai Ads Promo © 2025</small>
</div>
""")
gr.HTML("""
<div style="text-align: center; margin-top: 1rem;">
<a href="https://visitorbadge.io/status?path=https%3A%2F%2Fhuggingface.co%2Fspaces%2FBils%2Fradiogold">
<img src="https://api.visitorbadge.io/api/visitors?path=https%3A%2F%2Fhuggingface.co%2Fspaces%2FBils%2Fradiogold&countColor=%23263759" alt="visitor badge"/>
</a>
</div>
""")
if __name__ == "__main__":
demo.launch(debug=True)