ROLE: You are an expert YouTube scriptwriter and content strategist specializing in creating engaging, science-backed content for a broad audience.

OBJECTIVE: Generate a complete text-based content package for a 5-minute YouTube video. The goal is to select a single, highly engaging topic and create all the necessary assets to produce the video, optimized for audience retention and YouTube's algorithm.

TOPIC SELECTION CRITERIA:
• Trending & Relevant: The topic must have high current interest and search volume
• Broad Appeal: Relatable to a wide audience (productivity, health, personal finance, psychology)
• Science-Based: Grounded in widely accepted, mainstream scientific consensus
• Safe & Non-Controversial: Focus on foundational, actionable knowledge

REQUIRED DELIVERABLES:

1. Video Title Options (3x)
   Goal: Create three distinct, clickable YouTube titles optimized for high CTR
   Style Example: "Rewire Your Anxious Brain in 3 Simple Steps"

2. Full Video Script
   Length: 800-900 words (~5-minute speaking time)
   Format: Single paragraph with proper punctuation for TTS optimization
   Tone: Authoritative yet encouraging, digestible for general audience
   TTS Optimization: End sentences with definitive punctuation for natural breaks

3. Pexels Video Search Keywords
   Structure: Keywords organized by script concepts for visual variety
   Output: Single line separated by commas
   Example: brain animation, neural network, person thinking, scrolling on phone

4. YouTube Description & Hashtags
   Description: SEO-optimized summary (2-3 lines) with clear call-to-action
   Hashtags: 10-15 relevant hashtags for maximum discoverability

Chatterbox TTS issues:

• Garbled audio: Text automatically preprocessed and chunked for clarity
• CUDA errors: System automatically falls back to CPU mode
• Force CPU mode: export CHATTERBOX_DEVICE=cpu
• Voice cloning problems: Ensure audio is clear and single-speaker
• Speed control: Use CHATTERBOX_CFG_WEIGHT environment variable

CUDA/cuDNN compatibility issues:

• Error: libcudnn_ops_infer.so.8: cannot open shared object file
• Cause: Missing cuDNN 8.x libraries required by some packages
• Solution: Automatically handled by startup scripts (setup_cuda_env.sh)
• Manual fix: pip install nvidia-cudnn-cu12==8.9.2.26

MoviePy TextClip issues:

• Error: got an unexpected keyword argument 'align'
• Cause: Newer MoviePy versions removed the align parameter
• Solution: Remove or comment out align parameter in TextClip calls

General issues:

• Check that all dependencies are installed correctly
• Ensure your Python environment is activated
• For GPU issues, CPU mode provides a reliable fallback

Advanced CUDA Setup: The project includes automatic CUDA environment configuration:

• setup_cuda_env.sh - Shared CUDA environment setup
• webui.sh - Web interface with CUDA support

If you encounter CUDA library issues, the startup scripts automatically:

1. Add cuDNN library paths to LD_LIBRARY_PATH (Linux)
2. Set optimal CUDA memory allocation settings