Features
What you need to capture, organize, and review your meetings, all built in. No plugins, and nothing leaves your machine.
Core
What you get on day one. Detect a meeting, record it, transcribe it, summarize it, and find it again later.
Auto-detection
Watches running applications for Zoom, Teams, and Google Meet. Recording starts the moment a meeting begins and stops when it ends. You don't click anything.
Recording engine
A lightweight built-in recording engine that uses your GPU when one is available and falls back to software encoding when it isn't. Output is a standard .mp4 file at 1080p, 720p, or 480p. Nothing else to install.
Window-specific capture
Captures only the meeting window, not your whole screen. Private chats, emails, and other windows stay off the recording. Files live on your disk. No cloud, no telemetry.
Automatic transcription
When a recording finishes, autorec runs whisper.cpp on it locally and writes a plain-text transcript plus a subtitle (.srt) file. You can pick the model size you want; download it once and use it forever. Audio stays on your machine.
AI-powered summaries
Optional. Plug in your own API key for any OpenAI-compatible service (OpenAI, OpenRouter, Anthropic via a proxy, or a local llama.cpp server) and autorec generates titles and summaries. Only the transcript text is sent. Audio and video never leave the machine.
Built-in video library
Browse recordings as a card grid or a table, with auto-generated thumbnails. Search by title, sort by date or size. Recordings play in an embedded video player. No need to open anything else.
Transcript viewer and editor
View transcripts next to the video with clickable timestamps. Jump to any moment in the recording. Edit transcripts inline and save changes back to the .txt and .srt files on disk.
Trim and compress
Trim dead time off the start and end of a recording. Compress files to save disk space using preset percentages or a custom target size. Every job shows up in one processing queue.
System tray integration
Lives in your system tray, out of the way. Right-click to force-start or stop a recording, switch between window and full-screen capture, or open the recordings folder.
Configurable settings
Output directory, filename templates, quality presets, per-app toggles, transcription model, AI service configuration. All of it sits in a clean settings panel. Settings persist as plain JSON in your config directory, so you can read or edit them yourself.
Advanced and power-user controls
Eight more knobs for the people who want to tune autorec, automate around it, or wire it into a larger workflow.
Custom meeting watchers
Match any app or browser tab, not just the big three. Add your own watchers for Discord, Slack huddles, Whereby, Around, or any window or URL you want recorded.
For power users: Exact, glob, or regex patterns on window titles or browser URLs. Per-watcher priorities, plus JSON import/export so you can share watchers across machines or teams.
Pro audio controls
Pick separate input devices for your mic and your system audio. Adjust gain, mute, and noise suppression independently for each source. Visual level meters show what is actually being captured.
For power users: Per-source device hot-swap mid-recording, ±24 dB gain in 0.5 dB steps, RNNoise plus OS-level noise suppression chain. Falls back gracefully if a device disconnects.
Smart audio ducking
Quiets the other side automatically when someone is speaking, so the two halves of the conversation don't drown each other out. Tune it for two-way calls or podcast-style mic priority.
For power users: Two-way or mic-priority modes. Trigger on RMS envelope or RNNoise voice activity detection. Configurable threshold, attack, release, and depth.
Apps-only desktop audio
Record audio from specific applications instead of every sound on your desktop. Music, notifications, and other windows stay out of the meeting recording.
For power users: Per-process selection on Linux (PipeWire/PulseAudio sink routing) and Windows (WASAPI loopback). Re-scans for newly running apps on demand.
Hardware encoder acceleration
Uses your GPU when available, so recording barely touches your CPU. If the hardware encoder isn't available, or stops working mid-recording, the engine swaps to software automatically.
For power users: NVENC → QuickSync → VideoToolbox → libopenh264 fallback chain, with a graceful mid-recording swap if the GPU encoder drops out. All paths are LGPL-clean.
Webhooks and automation
POST a notification (and optionally the file itself) to any URL when a recording, transcript, or summary finishes. Wire autorec into Notion, Obsidian, Zapier, n8n, or your own scripts.
For power users: HTTP multipart or SFTP upload, custom JSON templates with field interpolation, retry with exponential backoff, and per-event chaining (e.g. fire webhook B only after webhook A succeeds).
Unified processing queue
Every transcription, model download, trim, compress, and summary in one place. Cancel running jobs, retry the ones that failed, and run bulk actions across many recordings.
For power users: Task history persists across app restarts (~/.local/share/autorec/tasks.json). Per-task cancel cleans up subprocesses and temp files. Sidebar badge shows active task count.
Pause, resume, and manual override
Pause a recording mid-meeting and resume when you're back. Force-start or force-stop from the tray when auto-detection isn't doing what you want.
For power users: Paused time is excluded from the elapsed counter. Tray menu offers force-start/stop and switching between window and full-screen capture. Audio device hot-swap also works while recording.
See it in action
A short walkthrough of the full flow: detect, record, transcribe, library.