This guide covers how to install and use the AI YouTube Shorts Generator on systems without an NVIDIA GPU. CPU mode supports all features but with slower transcription speeds.
CPU transcription is 5-10x slower than GPU mode. A 5-minute video may take 3-5 minutes to transcribe on CPU vs. 30 seconds on GPU.
System Requirements
CPU : Modern multi-core processor (Intel i5/AMD Ryzen 5 or better)RAM : 8GB minimum, 16GB recommendedPython : 3.10 or higherStorage : 2GB for dependencies + space for video filesOS : Windows 10/11, Ubuntu 20.04+, or macOS 11+ CPU mode requires significantly more RAM than GPU mode. Ensure you have at least 8GB available, especially for longer videos.
Windows CPU Installation
Install FFmpeg
Option 1: Using Chocolatey (Recommended) # Run PowerShell as Administrator choco install ffmpeg - y
Option 2: Using Scoop Option 3: Manual Installation
Install ImageMagick
choco install imagemagick - y
Configure ImageMagick policy: Edit C:\Program Files\ImageMagick-7.x.x-Q16-HDRI\config\policy.xml Find: < policy domain = "path" rights = "none" pattern = "@*" />
Change to: < policy domain = "path" rights = "read|write" pattern = "@*" />
Create Virtual Environment
# Navigate to project directory cd AI - Youtube - Shorts - Generator # Create virtual environment python - m venv venv # Activate virtual environment .\venv\Scripts\Activate.ps1
If you get an execution policy error, run: Set-ExecutionPolicy - ExecutionPolicy RemoteSigned - Scope CurrentUser
Install CPU PyTorch
IMPORTANT: Install PyTorch first with CPU support: pip install torch -- index - url https: // download.pytorch.org / whl / cpu
Install Other Dependencies
pip install - r requirements - cpu.txt
Configure API Key
Create a .env file in the project root: OPENAI_API = your_openai_api_key_here
Linux CPU Installation
Install System Dependencies
Ubuntu/Debian: sudo apt update sudo apt install -y ffmpeg libavdevice-dev libavfilter-dev \ libopus-dev libvpx-dev pkg-config libsrtp2-dev imagemagick
Fedora: sudo dnf install ffmpeg ffmpeg-devel imagemagick
Fix ImageMagick Policy
sudo sed -i 's/rights="none" pattern="@\*"/rights="read|write" pattern="@*"/' /etc/ImageMagick-6/policy.xml
On some systems, the policy file may be at /etc/ImageMagick-7/policy.xml. Check which version you have.
Create Virtual Environment
cd AI-Youtube-Shorts-Generator python3.10 -m venv venv source venv/bin/activate
Install CPU PyTorch
pip install torch --index-url https://download.pytorch.org/whl/cpu
Install Other Dependencies
pip install -r requirements-cpu.txt
Configure API Key
echo 'OPENAI_API=your_openai_api_key_here' > .env
macOS CPU Installation
Install Homebrew Dependencies
brew install ffmpeg imagemagick
Create Virtual Environment
cd AI-Youtube-Shorts-Generator python3.10 -m venv venv source venv/bin/activate
Install CPU PyTorch
pip install torch --index-url https://download.pytorch.org/whl/cpu
Install Other Dependencies
pip install -r requirements-cpu.txt
Configure API Key
echo 'OPENAI_API=your_openai_api_key_here' > .env
macOS with Apple Silicon (M1/M2/M3) can use MPS (Metal Performance Shaders) for some acceleration, but this configuration uses CPU mode for maximum compatibility.
Verify Installation Confirm CPU mode is active:
python -c "import torch; print('CUDA available:', torch.cuda.is_available())"
Expected output: If this shows True, you accidentally installed the GPU version of PyTorch. Uninstall and reinstall from the CPU index.
CPU Requirements File The requirements-cpu.txt file excludes all NVIDIA CUDA packages that are in the standard requirements.txt. Key differences:
What's Different in requirements-cpu.txt
Excluded packages:
All nvidia-* packages (CUDA runtime, cuDNN, etc.)
GPU-specific PyTorch wheels
Included packages:
CPU-only versions of core dependencies
faster-whisper==1.0.1 - CPU-compatible transcriptiontorch (installed separately from CPU index)All other standard dependencies (OpenCV, MoviePy, LangChain, etc.)
Total size: ~800MB of Python packages (vs. ~4GB with CUDA)Full Requirements List The requirements-cpu.txt includes:
# Key packages faster-whisper==1.0.1 # Speech-to-text transcription langchain==0.3.25 # LLM framework langchain-openai==0.3.0 # OpenAI integration moviepy==1.0.3 # Video editing opencv-python==4.8.1.78 # Computer vision pytubefix==9.1.1 # YouTube downloads python-dotenv==1.0.1 # Environment variables # And 70+ supporting dependencies...
PyTorch must be installed separately from the CPU index before installing requirements-cpu.txt.
Typical processing times for a 5-minute YouTube video:
Task GPU (CUDA) CPU Download 10-30s 10-30s Audio extraction 2-5s 2-5s Transcription 30-60s 3-5 min Highlight selection 5-10s 5-10s Video processing 20-40s 25-50s Total 1-2 min 4-7 min
Longer videos (15-30 minutes) can take 15-30 minutes to transcribe on CPU. Consider processing overnight for large batches.
Running in CPU Mode Usage is identical to GPU mode:
Interactive Mode Command-Line Mode python main.py "https://youtu.be/VIDEO_ID"
Batch Processing xargs -a urls.txt -I {} python main.py --auto-approve {}
The application automatically detects CPU mode and adjusts accordingly. No code changes are needed.
Optimization Tips
Close Background Applications
Free up RAM by closing unnecessary programs, especially browsers with many tabs.
Process Shorter Videos First
Start with 2-5 minute videos to get faster results. The highlight selection works best with 5-15 minute source videos anyway.
Use Batch Processing Overnight
For multiple videos, use the --auto-approve flag and run overnight: xargs -a urls.txt -I {} sh -c 'python main.py --auto-approve {} && sleep 10'
Monitor Resource Usage
Watch CPU and memory usage: Linux/macOS: Windows:
Open Task Manager (Ctrl+Shift+Esc)Switching Between CPU and GPU If you later add an NVIDIA GPU:
Install CUDA Toolkit
Follow NVIDIA’s installation guide for your OS.
Install GPU PyTorch
pip install torch --index-url https://download.pytorch.org/whl/cu118
Install Full Requirements
pip install -r requirements.txt
To revert to CPU mode:
Uninstall All Dependencies
pip uninstall -r requirements.txt -y
Install CPU PyTorch
pip install torch --index-url https://download.pytorch.org/whl/cpu
Install CPU Requirements
pip install -r requirements-cpu.txt
Troubleshooting CPU Installation Issue: Still Using GPU Libraries If you see CUDA-related errors after CPU installation:
# Verify no NVIDIA packages are installed pip list | grep nvidia
Should return nothing. If packages are listed:
pip uninstall $( pip list | grep nvidia | awk '{print $1}' ) -y
Issue: Out of Memory Errors If transcription fails with memory errors:
Close other applications to free RAMProcess shorter video segments (< 5 minutes)Increase system swap space (Linux) or virtual memory (Windows)
If CPU transcription is taking excessively long:
Check CPU usage : Should be near 100% during transcriptionDisable power saving : Ensure your system isn’t throttling CPUConsider alternatives : The AI Clipping API offers faster cloud processing
For occasional use, CPU mode works fine. For frequent processing or longer videos, GPU acceleration or a cloud API may be more practical.
When to Use CPU Mode CPU mode is ideal for:
💻 Laptops without dedicated GPUs
🖥️ Desktop systems with AMD or Intel graphics
🍎 Apple Silicon Macs (M1/M2/M3)
📦 Minimal installation footprint (~800MB vs. ~4GB)
🧪 Testing and development
Consider GPU mode for:
⚡ Frequent video processing
📹 Long-form content (15-30 minute videos)
🏭 Batch processing many videos
⏱️ Time-sensitive workflows
All features work identically in both modes. The only difference is processing speed.
Alternative Solutions If CPU performance is insufficient:
AI Clipping API : muapi.ai/playground/ai-clipping
No installation required
Faster processing
Better highlight selection algorithms
Google Colab : Run the tool on free GPU instances
Requires adapting the code for notebook format
Free tier has usage limits
Cloud VM with GPU : AWS, GCP, or Azure instances
Rent by the hour
Use the standard GPU installation
