Montaigne

A Python toolkit for presentation animation. Extract slides, translate visuals, generate voiceovers, and create videos—powered by Google Gemini AI, ElevenLabs, and local TTS.

Features

• PDF Extraction: Convert PDF presentations to high-quality images. Configurable DPI settings (150-300+) with PNG or JPG output formats.
• Script Generation: Generate professional voiceover scripts using a two-pass AI approach. Holistic context analysis, narrative arc awareness, and production notes with pronunciation guides.
• Image Translation: Translate text in images to any target language. Context-aware translations powered by Gemini.
• Voice Synthesis: Generate natural voiceover audio from scripts. Three providers: Gemini TTS (cloud), ElevenLabs (premium voices), or Coqui XTTS-v2 (local, no API key required).
• Video Generation: Combine translated slides and voiceover audio into polished videos. Configurable resolution up to 1920x1080.
• PowerPoint Export: Create PPTX presentations from PDF or images. Optionally add voiceover scripts as speaker notes for each slide.
• Cloud Deployment: Offload video generation to Google Cloud Run. Upload PDFs, process in the cloud, and download results with secure signed URLs.
• Model Configuration: Customize Gemini models for each operation. Use --model flags to switch between flash and pro models based on your needs.
• Video Annotation: Frame-accurate video and audio annotation tool. Add timestamps, export to WebVTT/SRT formats for captions. Waveform visualization with click-to-seek.
• Web Editor: Streamlit-based slide editor for managing presentations

Installation

Using pip

pip install montaigne

With optional dependencies

# Install with web editor support
pip install "montaigne[edit]"

# Install with annotation tool support
pip install "montaigne[annotate]"

# Install all optional dependencies
pip install "montaigne[all]"

Using uv

uv pip install montaigne

Using uvx (no installation required)

uvx --from montaigne essai setup
uvx --from montaigne essai script --input presentation.pdf

Setup

1. Get a Gemini API key from Google AI Studio
2. Create a .env file:

   GEMINI_API_KEY=your-api-key
   

3. Verify setup:

   essai setup
   

Usage

Extract PDF to Images

essai pdf presentation.pdf
essai pdf presentation.pdf --dpi 200 --format jpg

Generate Voiceover Script from Slides

essai script --input presentation.pdf
essai script --input slides_images/ --context "AI workshop"
essai script --input presentation.pdf --output custom_script.md
essai script --input presentation.pdf --model gemini-2.5-flash

Options:

• --input, -i: PDF file or folder of slide images
• --output, -o: Output markdown file path
• --context, -c: Additional context to guide script generation
• --model, -m: Gemini model to use (default: gemini-3-pro-preview)

Generate Audio from Script

essai audio --script voiceover.md
essai audio --script voiceover.md --voice Kore
essai audio --script voiceover.md --model gemini-2.5-flash-preview-tts

TTS Providers:

Gemini voices: Puck, Charon, Kore, Fenrir, Aoede, Orus

Local TTS with Coqui:

# Install Coqui dependencies
pip install "montaigne[coqui]"

# Generate audio locally (no API key required)
essai audio --script voiceover.md --provider coqui
essai audio --script voiceover.md --provider coqui --voice male
essai audio --list-voices --provider coqui

Coqui voices: female, male, neutral

Note: First run downloads the XTTS-v2 model (~1.5GB). Requires accepting the CPML license.

Options:

• --script, -s: Path to voiceover markdown script
• --provider, -p: TTS provider (gemini, elevenlabs, coqui)
• --voice, -v: TTS voice to use (default: Orus for Gemini, female for Coqui)
• --model, -m: Gemini TTS model (default: gemini-2.5-pro-preview-tts)

Translate Images

essai translate --input slides/
essai translate --input image.png --lang Spanish
essai translate --input slides/ --model gemini-2.0-flash-exp

Options:

• --input, -i: Image file or folder of images
• --lang, -l: Target language (default: French)
• --model, -m: Gemini model (default: gemini-3-pro-image-preview)

Create PowerPoint from PDF or Images

essai ppt --input presentation.pdf
essai ppt --input slides/ --script voiceover.md
essai ppt --input presentation.pdf --keep-images

This will create a .pptx file with each PDF page or image as a slide. If a voiceover script is provided, it will be added as speaker notes.

Generate Video from Slides

essai video --pdf presentation.pdf
essai video --images slides/ --audio audio/

Full Localization Pipeline

essai localize --pdf presentation.pdf --script voiceover.md --lang French

This will:

1. Extract PDF pages to images
2. Translate all images to the target language
3. Generate audio for all slides

Video/Audio Annotation Tool

Launch an interactive web UI for annotating videos or audio files with frame-accurate timestamps:

# Install annotation dependencies first
pip install "montaigne[annotate]"

# Launch annotation UI
essai annotate video.mp4
essai annotate audio.wav
essai annotate                        # Auto-detect media in current dir
essai annotate video.mp4 --network    # Make accessible on local network

# Export annotations
essai annotate video.mp4 --export srt   # Export to SRT (Premiere, DaVinci)
essai annotate video.mp4 --export vtt   # Export to WebVTT (browsers)
essai annotate video.mp4 --export json  # Export to JSON

Keyboard shortcuts:

Features:

• Frame-accurate timing using requestVideoFrameCallback API
• Waveform visualization with click-to-seek
• Light/dark theme toggle
• Local-first SQLite storage (zero-latency)
• Export to WebVTT, SRT, JSON formats

Web Editor

Launch a Streamlit-based web interface for managing slides and scripts:

# Install editor dependencies first
pip install "montaigne[edit]"

# Launch the editor
essai edit
essai edit --pdf presentation.pdf --script voiceover.md

Model Configuration

Each AI command supports a --model / -m flag to override the default Gemini model:

List available models:

essai models

Voiceover Script Format

Scripts should follow this markdown format:

## SLIDE 1: Title
**[Duration: ~45 seconds]**

Your narration text for slide 1 goes here.

---

## SLIDE 2: Next Topic
**[Duration: ~60 seconds]**

Narration for slide 2.

Demo

See the demo/hamlet/ folder for a complete example with:

• Sample PDF presentation
• Voiceover script
• Image asset

cd demo/hamlet
essai localize --lang French

Requirements

• Python 3.10+
• Google Gemini API key
• ffmpeg (for video generation)
• Dependencies: google-genai, python-dotenv, pymupdf, python-pptx, Pillow

Optional Dependencies

• edit: streamlit - Web editor interface
• annotate: flask - Video/audio annotation tool
• coqui: TTS, torch, torchaudio - Local TTS with Coqui XTTS-v2 (no API key required)
• cloud: fastapi, uvicorn, google-cloud-storage - Cloud API deployment