Auto-Cut to Beat, Speech-to-Text captions, Auto-Reframe for vertical delivery, Voice Isolate for outdoor vows, Optical Flow speed remap, and planar tracking — all local, all editable, all included.
Drag clips and a song onto the timeline — EDITBX detects beats and lays down a beat-cut sequence in seconds. Reel-ready, wedding-ready.
Built-in transcription workflow. Drop a clip, get editable burn-in captions — no upload, no API key.
Reframe 16:9 to 9:16, 1:1, 4:5, or 2.39:1 cinemascope with subject tracking. One-click vertical delivery.
Optical-flow slow motion creates intermediate frames where the camera left gaps. Ideal for silky bridal-walk re-times.
Dedicated adjustment-layer track — drop a grade, a look, or a vignette and it cascades over everything below.
Point and planar tracking for masks, callouts, and signage replacement — no external round-trip.
Every feature below is wired in the current codebase. Status badges reflect real build capabilities — not marketing fluff. Click any card to preview the artwork.
Hybrid chat assistant: built-in fast rules + optional Ollama / LM Studio localhost models. Suggests edits for wedding highlights, reels, export presets, scene/beat markers, color, audio cleanup, and project health — always with Apply + confirm before anything touches your timeline.
Ollama · LM Studio · Built-in GGUF · Offline rules
Generate caption cues from clip audio using macOS Speech (on-device) or bundled whisper.cpp (ggml-base.en.bin). Writes SRT sidecars and timeline caption tracks. Also powers the AI "Create subtitle" and Social Caption quick commands.
whisper.cpp · macOS Speech · SRT sidecar
Transcribe a clip with word-level timings, open the Transcript panel, select words, and press Delete to ripple-cut those sections from the timeline — like editing a document, not scrubbing frames.
Whisper word timings → Transcript panel → ripple extract
Find shot changes in the selected clip and split at each detected scene boundary. Uses FFmpeg scdet plus OpenCV refinement (or color-stats heuristic when OpenCV is unavailable).
FFmpeg scdet + OpenCV · color-stats fallback
Reframe the active sequence to vertical 9:16 for Reels/Shorts/TikTok. With OpenCV: scans people across each clip and writes smoothed transform keyframes. Without OpenCV: center-weighted heuristic crop (honestly labeled).
OpenCV person scan + SmartReframeApply · center heuristic fallback
Detect or segment subjects at the playhead and create follow masks for color grading or Fusion. ONNX U²-Net matte when model is present; OpenCV GrabCut fallback; Haar/HOG detect for Fusion mask nodes.
ONNX u2netp.onnx · GrabCut · Haar/HOG → Fusion nodes
Analyze a representative frame and apply balanced exposure, contrast, and color to the selected clip. Multi-zone luma/RGB statistics with adaptive local contrast — no cloud ML, but genuinely useful one-click grade starting point.
Multi-zone luma/RGB stats → clip grade
Build a music-driven montage that cuts on detected beats. Uses FFmpeg/DSP onset and energy analysis (not deep learning). Opens the Auto Beat Cut builder with beat markers you can refine.
FFmpeg/DSP beat analysis → cut planner
Upscale export resolution 2×–4× with bilateral denoise and sharpen pass on Deliver. Set Super Scale in Project Settings; applied at export time — preview stays fast, export gets finishing quality.
Project Super Scale · bilateral denoise · sharpen on export
AI assistant generates structured wedding highlight plans (Ceremony, Reception, First Dance…) and beat-aware vertical reels cut plans from your project context. Apply creates timeline markers — plans are suggestions, not silent auto-edits.
AiStructuredActionParser · WeddingDashboardPanel · marker Apply
Faster organization, caption drafts, beat hints, and practical suggestions — assistive tools that speed up real sessions.
Explore AI toolsEDITBX presents AI as a controlled assistant layer: analyze, suggest, review, apply. No hidden auto-editing, no confusing magic buttons.
Suggests timeline actions, cleanup steps, captions, transitions, and delivery checks without taking control away from the editor.
Analyzes music energy, beats, voice onsets, and markers to help build highlight edits and social reels faster.
Finds visual cut candidates for long footage, ceremony edits, documentaries, and imported reels.
Supports caption drafts through SRT or speech-to-text workflows so editors can review, correct, and publish clean subtitles.
Uses vision-assisted detection paths for face or people-aware masks, useful for portrait polish and vertical reframing.
Checks project, cache, proxy, RAM, decoder, and delivery states before final output so client exports stay predictable.
A compact workflow editors can understand immediately inside the website.
Import ceremony, reception, speeches, and reel footage
Analyze beats, scene cuts, markers, and dialogue-heavy ranges
Create editable suggestions for highlights, chapters, captions, and reels
Review everything on the timeline before applying
Export client-ready deliverables with safety checks
We only put a feature on the website once it's running in our internal builds.
Queue 9:16 Reel, 1:1, 4:5, and 16:9 master from a single timeline. Coming in the next round.
Seat sharing, audit log, and remote deactivation for studios with more than 3 workstations.
