AutoShorts is a local-first desktop application for turning long-form video or audio recordings into high-impact, vertical short-form clip candidates (9:16 portrait) with AI-powered viral moment ranking.
This repository implements the desktop app foundation using Tauri 2 + React + TSX + Rust + SQLite.
- Dynamic Multi-LLM Support: Supports both DeepSeek (default) and Claude (anthropic) for viral moment detection and hooks analysis.
- Automated Pipeline: Imports media, extracts audio, transcribes using Deepgram, and automatically analyzes and ranks moments in a single automated chain.
- Local SQLite Storage: Saves transcripts, candidates, custom names, and rendering data locally.
- Native Project Manager: Create, open, rename, and delete projects from the dashboard.
- Portrait Auto-Cropping: Automatically center-crops landscape videos to vertical H.264 portrait clips using native
ffmpegintegration. - Key Warnings: Built-in visual warnings that identify missing environment variables and prompt you directly in the UI.
To run the application, FFmpeg & FFprobe must be installed and available on your system PATH to handle cropping, audio extraction, and dynamic captions:
- macOS: Install using Homebrew:
Note: To ensure full captions rendering support, if standard Homebrew FFmpeg lacks drawtext/subtitles filters, tap and install the
brew install ffmpeg
homebrew-ffmpegformula:brew tap homebrew-ffmpeg/ffmpeg brew install homebrew-ffmpeg/ffmpeg/ffmpeg
- Windows: Install using Winget (in PowerShell):
(Or download the release build from gyan.dev and add it to your system PATH environment variables).
winget install Gyan.FFmpeg
- Linux: Install via your native package manager:
sudo apt install ffmpeg # Debian/Ubuntu sudo pacman -S ffmpeg # Arch Linux sudo dnf install ffmpeg # Fedora
Download the correct package matching your system from the latest GitHub Releases.
- Download:
- Apple Silicon (M1/M2/M3): Select the
aarch64.dmgpackage. - Intel Mac: Select the
x64.dmgpackage.
- Apple Silicon (M1/M2/M3): Select the
- Install: Double-click the
.dmgfile and drag AutoShorts to your Applications folder. - Bypass Gatekeeper (For unsigned local builds):
- Right-click
AutoShorts.appin Finder, select Open, and click Open in the warning dialog. - Alternatively, run this command in Terminal:
xattr -cr /Applications/AutoShorts.app
- Right-click
- Download: Select the
.msi(installer) or.exe(portable executable) package. - Install: Double-click the
.msifile to run the setup wizard. - SmartScreen Bypass: Since the package is self-signed, Windows SmartScreen may show a warning. Click "More Info" in the window and choose "Run anyway".
- Download: Select the
.deb(Debian/Ubuntu) or.AppImage(universal portable binary). - Install
.deb:sudo dpkg -i autoshorts_*.deb - Run
.AppImage: Make it executable and launch it:chmod +x autoshorts_*.AppImage ./autoshorts_*.AppImage
When you first launch the application, you will be greeted by an Onboarding Wizard that lets you choose your preferred workflow:
- Ollama Setup: Select a local model card (
llama3.2 3B,qwen2.5 3B, orqwen2.5 7B). The application will check if Ollama is running and automatically pull the model weights, showing a downloader progress bar. - Local Whisper: Follow the prompt instructions to verify Python is installed and run
pip3 install openai-whisperto enable fully offline transcription.
- Enter your API credentials for:
- Deepgram: For fast, accurate cloud transcription.
- DeepSeek: (Highly Recommended) For cheap, high-quality cloud moment detection.
- Claude: For premium copywriting, hooks, and moment detection.
- Click Save & Start to immediately load the dashboard.
- Update Credentials: Click the API Settings gear icon in the top right of your app dashboard to switch engines, select different local models, or update API keys.
- Reset Onboarding: If you want to switch from Cloud to Offline (or vice-versa) and start setup from scratch, click the Reset App Configuration & Onboarding button at the bottom of the API Settings panel.
Tip
LLM Provider Recommendation (Local vs. Cloud):
- Local Models (Ollama): While AutoShorts supports fully offline moments analysis via local Ollama models (like LLaMA 3.2 3B or Qwen 2.5 3B/7B), local models are generally not recommended for viral moment detection. Smaller 3B/7B models lack the context reasoning and mathematical capabilities needed to evaluate long transcripts and calculate accurate segment timestamps (often outputting fragments that are too short).
- DeepSeek (Highly Recommended): We strongly suggest using DeepSeek for moment detection. It offers top-tier reasoning capabilities (matching GPT-4/Claude 3.5 Sonnet) at a fraction of a cent per run (under $0.001 per transcript). You can get an API key instantly at platform.deepseek.com.
- Claude (Premium Option): Claude 3.5 Sonnet provides the absolute best hooks copywriting and emotional resonance, but is slightly more expensive than DeepSeek (typically $0.01 – $0.05 per run).
Copy .env.example to .env in the root folder:
cp .env.example .envFill in your API Keys:
DEEPGRAM_API_KEY=your-deepgram-api-key
DEEPSEEK_API_KEY=your-deepseek-api-key
ANTHROPIC_API_KEY=your-anthropic-api-key
# Choose your default AI analysis provider ("deepseek" or "claude")
LLM_PROVIDER=deepseekTo start the live-reloaded frontend and backend development shell:
npm install
npm run tauri:devTo build and package the native macOS app bundle (.app and .dmg installer):
npm run tauri:buildThe output installers will be built under src-tauri/target/release/bundle/.
If AutoShorts helps you create content faster, consider supporting its development.
Your support helps fund new features, bug fixes, and ongoing improvements.
👉 https://polar.sh/checkout/polar_c_eZfQSAesVTAhaNyDtC8GnzySlU1yqflU62wwg2EfFDF