tmoroney/auto-subs

What it solves

AutoSubs is a local-first AI subtitle generator that allows users to transcribe audio and video files without sending data to the cloud. It eliminates the need for expensive subscriptions and ensures privacy by processing all transcription and translation tasks on-device.

How it works

The application leverages several local AI models for speech-to-text and speaker identification. It uses whisper-rs for OpenAI's Whisper models and ONNX Runtime for a variety of other models including Moonshine, Parakeet, Canary, and Cohere. It also incorporates a Silero VAD model for voice activity detection and a dedicated speaker diarization model to identify and label different speakers.

Who it’s for

Content creators, video editors, and translators who need high-quality, private local transcription. It is specifically designed to integrate with professional video editing software like DaVinci Resolve, Adobe Premiere Pro, and Adobe After Effects.

Highlights

• Local-First Processing: All transcription and translation happen on-device, ensuring no data leaves the machine.
• Broad Model Support: Supports a wide array of models (Whisper, Moonshine, Parakeet, SenseVoice, Canary, Cohere, GigaAM) for different languages and accuracy levels.
• Professional Integrations: Direct connection to DaVinci Resolve and CEP extensions for Adobe Premiere Pro and After Effects.
• Speaker Diarization: Ability to identify and label different speakers for per-speaker styling.
• Cross-Platform: Available for macOS, Windows, and Linux.