Pro NLE Architecture Update
A robust foundation for our macOS video editor, featuring a multi-clip NLE architecture, precise per-overlay transcription, and synchronized recording.
Pro-Grade NLE Architecture
We are laying down a robust foundation for macOS video editing. Version 3.1.2 introduces the core NLE (Non-Linear Editing) architecture. This shared timeline foundation establishes a true multi-clip timeline. It utilizes a “proxy-everything” approach with a dedicated frame extraction server, ensuring that your video creation workflow remains incredibly responsive, even with complex multi-clip NLE arrangements.
The timeline now natively supports waveform visualization and precise volume controls across fully refactored media track components, giving you complete command over your project’s pacing and audio.
Advanced Transcription Capabilities
Automated captions with local AI now support per-overlay transcription. Trim-aware extraction targets individual video clips, with concurrent guards and a new TranscriptionSlice store producing TextOverlay outputs.
We’ve also added an intuitive OverlayCaptionsTrack phrase sub-row directly below your selected video overlay, making it simpler than ever to refine your AI captions. Say goodbye to wasted processing power: active transcription is now automatically canceled when you hide the window.
Recording & Processing Refinements
We also optimized the recording engine:
- Flawless Synchronization: We’ve completely synchronized webcam and microphone recording, actively preventing dropped frames and ensuring your audio and video are perfectly locked.
- Improved H.264 Encoding: We’ve implemented IOSurface-backed H.264 encoding and a strict first-frame readiness gate, alongside recording pre-caching during countdowns. This guarantees instantly usable footage from the moment you hit record.
- Accurate Silence Removal: We now rigidly clamp our silence remover segments directly to video clip boundaries. This thoroughly prevents any visual overlap caused by audio overruns.
- Source Readiness Validation: We’ve improved source title re-resolution utilizing custom React hooks and implemented timeout protections with final output validation, securing your captures.
Engine Polish
Behind the scenes, we’ve removed stale type-guards in the transcribeOverlay workflow, vastly simplified storage loops down to raw-words per-overlay, and updated our Rust FFI integrations to cleanly handle explicit null parameters, ensuring optimal system stability.