app.py

import streamlit as st
from PIL import Image
import numpy as np
import io
import base64
import time
import os

# Disable Streamlit config file creation - Critical for Hugging Face Spaces
os.environ["STREAMLIT_GLOBAL_DEVELOPMENT_MODE"] = "false"
os.environ["STREAMLIT_GLOBAL_RECURSION"] = "false"

# Page setup - Vertical layout for 9:16 aspect
st.set_page_config(
    page_title="Free Hedra Clone",
    page_icon="🤖",
    layout="centered",
    initial_sidebar_state="expanded"
)

st.title("🎭 Free AI Avatar Generator")
st.subheader("Audio-Driven Speaking Avatar (9:16 Portrait)")

# Initialize session state
if "frame" not in st.session_state:
    st.session_state.frame = 0
if "avatar_img" not in st.session_state:
    st.session_state.avatar_img = None
if "audio_data" not in st.session_state:
    st.session_state.audio_data = None
if "audio_duration" not in st.session_state:
    st.session_state.audio_duration = 5
if "mime_type" not in st.session_state:
    st.session_state.mime_type = "audio/mp3"

# Create default 9:16 avatar
def create_default_avatar():
    width, height = 720, 1280
    img = Image.new('RGB', (width, height), color=(73, 109, 137))
    return img

# Sidebar configuration
with st.sidebar:
    st.header("🎛️ Configuration")
    
    # Audio file upload
    audio_file = st.file_uploader(
        "Upload Audio File:",
        type=["mp3", "wav"],
        help="Upload an audio file (MP3 or WAV)"
    )
    
    # Process audio immediately after upload
    if audio_file:
        st.session_state.audio_data = audio_file.read()
        if audio_file.name.endswith('.wav'):
            st.session_state.mime_type = "audio/wav"
        else:
            st.session_state.mime_type = "audio/mp3"
        
        # Set fixed duration since we can't reliably calculate it
        st.session_state.audio_duration = 5
        st.info("Audio duration set to 5 seconds (fixed)")
    
    # Avatar image upload
    uploaded_image = st.file_uploader(
        "Upload Avatar Image:", 
        type=["png", "jpg", "jpeg"]
    )
    
    # Process image immediately after upload
    if uploaded_image:
        try:
            img_data = uploaded_image.read()
            st.session_state.avatar_img = Image.open(io.BytesIO(img_data))
        except:
            st.error("Error processing image")
    
    # Animation options
    st.markdown("---")
    st.markdown("### Animation Options")
    pulse_intensity = st.slider("Pulse Intensity", 0.1, 1.0, 0.3, 0.1)
    animation_speed = st.slider("Animation Speed", 0.1, 2.0, 0.5, 0.1)
    mouth_intensity = st.slider("Mouth Movement", 0.5, 2.0, 1.2, 0.1)
    custom_duration = st.slider("Animation Duration (seconds)", 1, 30, 5, 1)

# Create speaking animation
def animate_avatar(image, intensity=0.3, anim_speed=0.5, mouth_factor=1.2):
    try:
        # Convert to RGB if needed
        if image.mode != 'RGB':
            image = image.convert('RGB')
        
        img_array = np.array(image)
        height, width, _ = img_array.shape
        
        # Create pulsation effect
        pulse = 1.0 + intensity * np.sin(st.session_state.frame * anim_speed)
        
        # Apply stronger effect to bottom half
        for y in range(height):
            position_factor = y / height
            if position_factor > 0.6:  # Mouth region
                region_intensity = intensity * mouth_factor
                region_pulse = 1.0 + region_intensity * np.sin(st.session_state.frame * anim_speed * 1.5)
            else:
                region_pulse = pulse
                
            img_array[y] = np.clip(img_array[y] * region_pulse, 0, 255).astype(np.uint8)
        
        return Image.fromarray(img_array)
    except Exception as e:
        st.error(f"Animation error: {str(e)}")
        return image

# Audio player component
def audio_player(audio_bytes, mime_type="audio/mp3"):
    audio_base64 = base64.b64encode(audio_bytes).decode()
    audio_html = f"""
    <audio controls autoplay style="width: 100%">
        <source src="data:{mime_type};base64,{audio_base64}" type="{mime_type}">
    </audio>
    """
    st.markdown(audio_html, unsafe_allow_html=True)

# Create default avatar if needed
if st.session_state.avatar_img is None:
    st.session_state.avatar_img = create_default_avatar()

# Display avatar preview
st.subheader("Avatar Preview")
st.image(
    st.session_state.avatar_img, 
    caption="Your Avatar",
    use_column_width=True
)

# Generate button
if st.session_state.audio_data:
    if st.button("Generate Speaking Avatar", type="primary", use_container_width=True):
        # Create animation container
        st.subheader("Speaking Avatar")
        animation_placeholder = st.empty()
        
        # Play audio
        st.subheader("Audio Player")
        audio_player(st.session_state.audio_data, st.session_state.mime_type)
        
        # Create animation
        st.subheader("Avatar Animation")
        
        # Animate for custom duration
        start_time = time.time()
        frames = 0
        
        while time.time() - start_time < custom_duration:
            try:
                animated_img = animate_avatar(
                    st.session_state.avatar_img, 
                    pulse_intensity,
                    animation_speed,
                    mouth_intensity
                )
                animation_placeholder.image(animated_img, use_column_width=True)
                st.session_state.frame += 1
                frames += 1
                time.sleep(0.1)
            except Exception as e:
                st.error(f"Error during animation: {str(e)}")
                break
        
        # Download audio button
        st.subheader("Download Options")
        st.download_button(
            label="Download Audio",
            data=st.session_state.audio_data,
            file_name=f"avatar_audio.{'mp3' if st.session_state.mime_type == 'audio/mp3' else 'wav'}",
            mime=st.session_state.mime_type,
            use_container_width=True
        )
else:
    st.warning("Please upload an audio file to generate the speaking avatar")

# Footer
st.markdown("---")
st.caption("""
    **Free Hedra Clone** - Audio-driven avatar animation
    All processing done in memory - No file system writes
""")