Auto-Editor

Auto-Editor is the open-source AI video editing tool released by the WyattBlue team in September 2022, with ★3.5k+ on GitHub, licensed under GPL-3.0, and written primarily in Python (FFmpeg backend). By analyzing audio waveforms + subtitle density, it auto-identifies and cuts out silences, stutters, filler words (um / uh / like / etc.), and repeated sentences from videos, outputting tightly-paced final videos. CLI-first, with an av (beta) GUI desktop application. Suitable for Vloggers / podcasters / online course creators / live stream recording / meeting recordings for quick post-production refinement.

Core capabilities center on "audio-driven editing + modular thresholds". Audio analysis: librosa / custom analyzer extracts volume curves, marks silence (default threshold -30dB, duration ≥ 0.3s). Cut policies: configurable --edit audio:threshold=0.05 / --edit motion:threshold=0.02 / --edit subtitle:style=mute. Filler-word removal: built-in "filler word list" (English: um / uh / like / you know / I mean; Chinese needs custom regex). Subtitle-driven cutting: input SRT, auto-cut gaps between subtitles. Edit modes: audio (only cut silence, safe), motion (cut static shots, ideal for talking heads), speech (cut filler words), subtitle (split by subtitle sentences). Output: re-encoded MP4 / MKV / mov, preserving original video stream codec or transcoding to H.264 / H.265. Ecosystem: FFmpeg / PyAV backend; WASM browser version (beta); GUI av command based on Electron; Docker images available.

Use cases include: Vlogger / YouTuber post-production (cut stutter + silence), podcast video versions (cut ums / uhs), online course refinement, live stream recording compression, meeting recording slim-down, and talking-head short-video secondary refinement. Hardware floor: CPU only (analysis is single-core sufficient), GPU not required; long videos (1-2 hours) take 5-10 minutes to analyze; output encoding can leverage GPU acceleration (NVENC). Onboarding: pip install auto-editor → auto-editor input.mp4 --edit audio:threshold=0.05 -o output.mp4; GUI users auto-editor --gui (or av command).

Licensed under GPL-3.0, requires derivative source disclosure (if distributed) — ideal for personal / internal tool use; embedding in commercial SaaS requires care (can route around with ffmpeg commands + self-implemented logic). Positioning versus VisLane's already-listed descript (commercial AI editing + text-driven copilot): Auto-Editor is "CLI-level open-source core"; descript is "commercial SaaS full-feature" (text-driven editing + AI copilot Underlord + transcription + multi-cam). The two overlap heavily in the "auto-cut silence / stutter" scenario — Auto-Editor is free and embeddable in your own pipeline, descript has a better UX but starts at $24/month subscription. Auto-Editor also complements ShortGPT — ShortGPT does content generation, Auto-Editor does post-production refinement; Auto-Editor's GPL-3.0 doesn't conflict with ShortGPT's MIT (MIT doesn't restrict GPL dependency direction).