autoshorts README

The project's README from github.com/JayWebtech/autoshorts. As one markdown file: README.md.

⭐ Support AutoShorts

AutoShorts

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.

Screenshot 2026-06-22 at 4 10 13 PM

Key Features

Prerequisites

To run the application, FFmpeg & FFprobe must be installed and available on your system PATH to handle cropping, audio extraction, and dynamic captions:

Installation Guide (For Users)

Download the correct package matching your system from the latest GitHub Releases.

🖥️ macOS Installation

  1. Download:
    • Apple Silicon (M1/M2/M3): Select the aarch64.dmg package.
    • Intel Mac: Select the x64.dmg package.
  2. Install: Double-click the .dmg file and drag AutoShorts to your Applications folder.
  3. Bypass Gatekeeper (For unsigned local builds):
    • Right-click AutoShorts.app in Finder, select Open, and click Open in the warning dialog.
    • Alternatively, run this command in Terminal:
      xattr -cr /Applications/AutoShorts.app

🪟 Windows Installation

  1. Download: Select the .msi (installer) or .exe (portable executable) package.
  2. Install: Double-click the .msi file to run the setup wizard.
  3. SmartScreen Bypass: Since the package is self-signed, Windows SmartScreen may show a warning. Click "More Info" in the window and choose "Run anyway".

🐧 Linux Installation

  1. Download: Select the .deb (Debian/Ubuntu) or .AppImage (universal portable binary).
  2. Install .deb:
    sudo dpkg -i autoshorts_*.deb
  3. Run .AppImage: Make it executable and launch it:
    chmod +x autoshorts_*.AppImage
./autoshorts_*.AppImage

🚀 First-Launch Onboarding & AI Configuration

When you first launch the application, you will be greeted by an Onboarding Wizard that lets you choose your preferred workflow:

Option A: Fully Offline (Ollama + Whisper)

  1. Ollama Setup: Select a local model card (llama3.2 3B, qwen2.5 3B, or qwen2.5 7B). The application will check if Ollama is running and automatically pull the model weights, showing a downloader progress bar.
  2. Local Whisper: Follow the prompt instructions to verify Python is installed and run pip3 install openai-whisper to enable fully offline transcription.

Option B: Cloud API Keys

  1. 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.
  2. Click Save & Start to immediately load the dashboard.

⚙️ Modifying Settings & Resetting Onboarding

[!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).

Developer Guide

1. Setup Environment Configuration

Copy .env.example to .env in the root folder:

cp .env.example .env

Fill 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=deepseek

2. Run in Development Mode

To start the live-reloaded frontend and backend development shell:

npm install
npm run tauri:dev

3. Build the Application

To build and package the native macOS app bundle (.app and .dmg installer):

npm run tauri:build

The output installers will be built under src-tauri/target/release/bundle/.

❤️ Support AutoShorts

If AutoShorts helps you create content faster, consider supporting its development.

Your support helps fund new features, bug fixes, and ongoing improvements.

👉 https://buy.polar.sh/polar_cl_ZjNgejG1JnPQqVXyMmEmq7vpdwJUFBqx4qahw4BqBCP

autoshorts (JayWebtech/autoshorts)

Searching Database…

1.3K ·GitHub
Autoshorts
What is this?AutoShorts is a local-first desktop app that turns long-form video or audio recordings into vertical (9:16) short-form clip candidates, ranking the most viral moments with AI.
Is this a cloud service or does it run on my computer?It's a native desktop app built with Tauri 2, React and Rust, not a web service; it keeps transcripts, candidates and project data in a local SQLite database on your machine.
Do I need to install FFmpeg myself?Yes, FFmpeg and FFprobe must already be installed and available on your system PATH before running the app, since it uses them natively for cropping, audio extraction and caption rendering.
Can I run this fully offline without any AI accuracy trade-off?You can go fully offline with Ollama plus local Whisper, but the README itself warns that small 3B/7B local models aren't great at moment detection because they lack the reasoning needed to score long transcripts and timestamp segments accurately.
Does my recording ever leave my computer?Only if you choose the cloud workflow, where audio goes to Deepgram for transcription and transcripts go to DeepSeek or Claude for moment analysis; everything else, including transcripts and candidates, is saved locally in SQLite.
Which AI providers can I use for moment detection?AutoShorts supports DeepSeek (default, recommended, under $0.001 per transcript) and Claude (premium, $0.01-0.05 per run, best for hooks/copywriting), plus Deepgram for cloud transcription or local Whisper if you go offline.
Is the installer code-signed?No. Windows builds are self-signed, so SmartScreen will warn and you need to click 'Run anyway'; macOS builds are unsigned, so you must right-click the app and choose Open to bypass Gatekeeper.
What happens the first time I launch the app?An Onboarding Wizard lets you pick a fully offline workflow (Ollama local model plus local Whisper) or a cloud workflow (Deepgram, DeepSeek or Claude API keys); you can switch or reset this later from the API Settings panel.
What are similar projects?autoclip, pyvideotrans, meetily, VidBee, screenpipe, VideoHighlighter, openwhispr, NextChat
Which topics is it in?short-video, ai-video, transcription, social-media, whisper, speech-to-text, tauri, content-creation, video-processing, ffmpeg, local-llm, ollama
Creator
JayWebtech
Size
1.5 MB
Stars
1.3K
Updated
CommitOct 11Releasev0.1.5Jul 01
Age
3mo
Commits · 90 Days