AI-Powered Presentation Speech Generator - Transform your slides into natural, professional speech transcripts using advanced AI models.
- 📊 Smart Slide Analysis - Automatically parses PDF presentation content with vision AI
- 🎙️ Speech Rate Detection - Upload a 20-second audio sample for automatic speed calculation
- 🎭 Multiple Speech Styles - Supports lively, serious, motivational, educational, and conversational styles
- 🌐 Multilingual Support - Generate transcripts in 8 languages:
- 繁體中文 (Traditional Chinese)
- English
- 简体中文 (Simplified Chinese)
- 日本語 (Japanese)
- 한국어 (Korean)
- Español (Spanish)
- Français (French)
- Deutsch (German)
- 🤖 Multiple AI Models - Choose from GPT-5.1, o3, GPT-4o, or GPT-4o-mini
- 👨🏫 Expert Role Playing - AI generates content from specific expert perspectives
- 💡 Speech Technique Suggestions - Includes gestures, tone, and pause recommendations
- 📥 One-Click Download - Export transcripts in TXT format
This project provides two notebook versions for different users:
-
presentation_transcript_generator_zh-TW.ipynb- Traditional Chinese (繁體中文)- Code comments and docstrings in Traditional Chinese
- For Taiwanese and Chinese-speaking developers
-
presentation_transcript_generator_en.ipynb- English- Code comments and docstrings in English
- For international developers
Note: Both versions have identical functionality. All Chinese UI elements are preserved for the best user experience.
-
Open the notebook in Google Colab:
- Chinese version: Upload
presentation_transcript_generator_zh-TW.ipynb - English version: Upload
presentation_transcript_generator_en.ipynb - Open with Google Colab
- Chinese version: Upload
-
Set your OpenAI API Key:
- Use Colab Secrets (recommended): Add
GPT_API_KEYin Secrets - Or enter manually when prompted
- Use Colab Secrets (recommended): Add
-
Run all cells and follow the interactive UI
- Python 3.10 or higher
- FFmpeg (for audio processing)
- OpenAI API Key
-
Clone the repository
git clone https://github.com/yourusername/slidescript-ai.git cd slidescript-ai -
Install FFmpeg
macOS:
brew install ffmpeg
Ubuntu/Debian:
sudo apt-get update sudo apt-get install ffmpeg
Windows:
- Download from FFmpeg website
- Add to system PATH
-
Install Python dependencies
pip install -r requirements.txt
-
Run the Streamlit app
streamlit run app.py
-
Access the app
- Open your browser to
http://localhost:8501 - Enter your OpenAI API Key in the sidebar
- Start generating transcripts!
- Open your browser to
-
Upload PDF Slides
- Click "選擇 PDF 檔案" to upload your presentation
- Maximum file size: 50MB
- Supported format: PDF
-
Set Speech Duration
- Enter your target presentation length (1-180 minutes)
-
Configure Speech Rate
- Choose from preset speeds: Slow (150), Medium (200), Fast (250)
- Or upload a 20-second audio sample for automatic analysis
-
Select AI Model
- GPT-5.1 ⭐ (Recommended): Best multimodal understanding
- o3: Strong reasoning for complex logic
- GPT-4o: Balanced performance
- GPT-4o-mini: Fast and economical
-
Choose Speech Style
- Lively (活潑)
- Serious (嚴肅)
- Motivational (激勵)
- Educational (教學)
- Conversational (對話)
-
Fill in Details
- Topic: Your presentation subject
- Audience: Target audience (e.g., university students, professionals)
- Language: Output language (8 options available)
- Expert Role (Optional): AI will assume this expert identity
- Include Tips: Check to add speech technique suggestions
-
Generate & Download
- Click "生成逐字稿" to generate
- Review the transcript
- Download as TXT file
- 📚 Academic Presentations - Convert research slides into lecture scripts
- 💼 Business Pitches - Create professional presentation narratives
- 🎓 Educational Content - Transform teaching materials into engaging scripts
- 🌍 Multilingual Events - Generate transcripts in multiple languages
- 🎤 Conference Talks - Prepare speaker notes with timing guidance
| Model | Best For | Speed | Cost |
|---|---|---|---|
| GPT-5.1 ⭐ | Deep image-text analysis, multimodal understanding | Medium | High |
| o3 | Complex reasoning, logical analysis | Slow | High |
| GPT-4o | Balanced performance, general use | Fast | Medium |
| GPT-4o-mini | Quick results, basic needs | Very Fast | Low |
- Frontend: Streamlit (local) / IPyWidgets (Colab)
- PDF Processing: PyMuPDF (fitz)
- Audio Analysis: OpenAI Whisper-1 API
- AI Models: OpenAI GPT-5.1/o3/GPT-4o/GPT-4o-mini
- Audio Conversion: pydub + FFmpeg
- OpenAI API account with available credits
- Supported models: gpt-5.1, o3, gpt-4o, gpt-4o-mini
- Audio API: whisper-1
openai>=1.12.0
pymupdf>=1.23.0
pillow>=10.0.0
pydub>=0.25.1
streamlit>=1.31.0
- Python 3.10+
- FFmpeg (for audio processing)
- 4GB+ RAM recommended
- Internet connection for API calls
- API Keys: Never commit API keys to version control
- Data Privacy: All processing is done via OpenAI API (subject to their terms)
- Local Files: Temporary files are stored in
/tmp/and cleaned up after use - Best Practice: Use environment variables or secrets management for API keys
Contributions are welcome! Please feel free to submit a Pull Request.
- Fork the repository
- Create your feature branch (
git checkout -b feature/AmazingFeature) - Commit your changes (
git commit -m 'Add some AmazingFeature') - Push to the branch (
git push origin feature/AmazingFeature) - Open a Pull Request
This project is licensed under the MIT License - see the LICENSE file for details.
- Built with OpenAI API
- PDF processing powered by PyMuPDF
- Audio handling by pydub
- UI framework: Streamlit
If you encounter any issues or have questions:
- 🐛 Report a Bug
- 💡 Request a Feature
- 📧 Contact: geomingical@gmail.com
If you find this project helpful, please consider giving it a star! ⭐
Made with ❤️ by a passionate developer using top-tier programming principles and UX design
Last Updated: December 2025