The friendly multilateral video toolkit built for artists by artists. It's your best friend.

154 effect/utility programs · node-based GUI + scriptable CLI · 100% local — no cloud, no accounts, no API keys

---

Videobeaux is a video-processing toolkit built around one big library of effects — glitch/corruption looks, temporal/frame effects, colour and visual filters, compositing tools, transcript-driven editing, and AI-assisted captioning/narration — and two ways to drive it:

• The GUI: drag effect nodes onto a canvas, wire them together into a pipeline, hit Run.
• The CLI: python3 -m videobeaux.cli -P <program> -i input.mp4 -o output.mp4 [options], script it however you want.

Both sit on the exact same 155 programs — nothing is GUI-exclusive or CLI-exclusive except a couple of legacy modes noted below. Everything runs on your machine. The one genuinely optional exception — AI narration scripting via a local Ollama model — is opt-in and the app works completely fine without it.

Contents

• The GUI
  • For users — just want to run it?
  • What you get
  • Optional AI features
• The CLI
  • Basic usage
  • Examples
• Installing from source
• Building GUI installers
• Programs

The GUI

A visual, node-based editor for building video pipelines out of videobeaux's effects. Input node → chain of effect nodes → Output node, connected however you like — branch, fan-out, batch-process a whole folder. No cloud services, no accounts, no telemetry.

  

Left: the effect library, organized by category — drag anything onto the canvas. Right: a frame from Kinetic Captions, one of the AI & Advanced programs — the actively-spoken word pops larger in real time, rendered locally with no ML model.

For users — just want to run it?

Download the installer for your platform from the Releases page:

• macOS: download the .dmg, open it, drag videobeaux into Applications.
• Windows: download the .exe and run it.

No Python, Node, or ffmpeg install needed — everything the app needs to run is bundled inside.

On first launch, a Setup screen walks you through getting fully working:

• Confirms the bundled Python and ffmpeg are working correctly.
• Lets you download a speech-recognition model (needed for transcript-driven effects — Silence Xtraction, Word Xtraction, Grep Supercut, Ngrams, Qwikchop Deluxe, Auto Narrate) — pick small/fast or large/accurate, right from the app.

Reopen this screen anytime from the ⚙ Setup button in the header — it's not a one-time wizard.

A couple of things to expect on an unsigned build:

• macOS: Gatekeeper will say the app "can't be opened because it is from an unidentified developer." Right-click (or Control-click) the app and choose Open, then confirm — you only need to do this once. Proper code signing/notarization is a planned follow-up, not yet in place.
• Windows: SmartScreen may show a similar "Windows protected your PC" warning. Click More info → Run anyway.

What you get

• A node canvas — drag any of the 155 programs onto it, wire Input → effects → Output. Multiple effects chain into one pipeline; branch one output into several downstream effects.
• Batch mode — point an Input node at a folder instead of a file and the whole pipeline runs once per video. Programs that natively produce many files from one input (Qwikchop, Qwikchop Deluxe, Extract Frames) fan out automatically.
• Save/load presets — snapshot the whole node graph (including every field you've filled in) to a .vbpreset.json file and reload it later.
• Live model pickers — Vosk (speech-to-text), Ollama (local LLM), and kokoro-tts voice dropdowns are all populated by asking the actual installed tool what it has, not a hardcoded list — install a new model and it just shows up.
• A real-time colour picker on every colour field (captions, LUTs, watermarks) — a native OS colour wheel, synced with a plain hex field you can also just paste into.

Licensing & the bundled ffmpeg

Videobeaux runs ffmpeg as a separate program. The ffmpeg that installers bundle is built by this project from source (tools/ffmpeg-build/, CI in .github/workflows/ffmpeg-build.yml):

• configured --enable-gpl (for libx264 / libx265) and never --enable-nonfree — the build script, the CI jobs, the installer build and the app's own Setup check all refuse a nonfree ffmpeg, because such a build can't be redistributed;
• every release publishes the exact source archives of ffmpeg and each library next to the binaries, plus the configure line and a licence summary (FFMPEG-LICENSES.txt, also in LICENSES/);
• you can use your own ffmpeg instead: Setup → Use my own ffmpeg… (for example an LGPL build).

Which ffmpeg runs? The installed app always uses the one in its own resources/ffmpeg. npm run dev looks, in order, at your own choice (Setup), <repo>/.tools/ffmpeg (what Repair downloads), Homebrew's ffmpeg-full, then whatever is on PATH — Setup → Technical details shows the one in use, its version and licence flavor, and a ✓/✗ list of the features Videobeaux needs. That list lives in videobeaux/assets/ffmpeg_requirements.json and the same check is available from a terminal: python -m videobeaux.utils.ffmpeg_caps [--ffmpeg /path/to/ffmpeg]. Programs that need something the build lacks (for example AV1 encoding) stop with a plain-language message instead of a raw ffmpeg error. Our own build includes AV1 (SVT-AV1 encode, dav1d decode).

Build it yourself on macOS: tools/ffmpeg-build/build-macos.sh (≈15 min the first time, a few minutes afterwards; produces dist/ffmpeg, dist/ffprobe, the configure line and the source archives). Bundled AI models keep their own licences (see Setup → Local AI features). This section is an engineering summary, not legal advice — have your distribution model reviewed before selling.

Looks and canvas appearance

🎨 Appearance lets you try different Looks (Classic, Patterns, Signatures, Output art — see ui_design/looks-README.md; Classic is the original and always one click away), colour the connection lines (Ink / By module / Custom, with thickness), and put a JPEG/PNG/WebP behind the canvas with its own opacity. Every module can be folded (▾ or double-click its header; Controls → ▤ / ▦ fold or unfold all) and the fold state is saved in presets.

Canvas tips

• Connect with two clicks — click a dot, a line follows your cursor, click another dot. (Dragging still works; Esc or a click on empty space cancels.)
• Select several — ⌘/Ctrl-click, Shift-drag a box, or ⌘A; selected programs get a bold yellow ring and a banner with a Delete button. Backspace/Delete removes them all (undo with ⌘Z).
• Drop files in — drag videos from Finder / Explorer onto the canvas: each becomes an Input node. 📸 Snapshot saves a PNG of the whole patch together with its preset JSON. The 🔥 button (bottom-left) toggles one-step runs that overwrite without asking. Drag the console's top edge to resize it; Copy copies its text.
• Groups — select two or more programs and press ⌘G (or Group): they move together, can be locked, renamed and colored, and a caret folds the whole group into one compact card (⇧⌘G ungroups). Folding is only visual — connections and what runs are unchanged.
• Copy / paste — select programs, ⌘C, then ⌘V pastes them (with the connections between them) at your cursor; ⌘D duplicates in place. The small bar at the bottom-left has the same buttons and can be switched off under 🎨 Appearance.
• Number fields — type a value (the box can be emptied and retyped) or drag the slider; the faint strip under it is a good starting range, the tick is the default, ↺ resets, and many programs have Presets chips.
• Help, Clear, hide panels — ? (or F1) opens a short help; Clear empties the canvas after confirming (⌘Z undoes); ⌘B hides the program list and ⌘J the console (progress stays visible).
• Layout editor (Lagkage) — press ✎ next to Layout JSON to drag images, GIFs and videos onto a stand-in for your video instead of writing JSON. Positions are saved as percentages, so a layout works at any resolution. The same ✎ button mechanism (components/helpers/registry.js) can host editors for other programs.
• Transparent keying — Chroma Key / Luma Key with Background = transparent needs the Output node set to WEBM or MOV.

Add your own photobooth filters

Drop a .py file into ~/.videobeaux/filters/ (or the folder in VIDEOBEAUX_USER_FILTERS) and restart:

from videobeaux.utils.booth_filters import register

@register("My filters", "Invert red")      # shows up in Filter Library → Filter as "My filters · Invert red"
def invert_red(img, ctx):                  # img: BGR uint8 frame; ctx.t = seconds, ctx.faces = [(x, y, w, h)]
    out = img.copy()
    out[..., 2] = 255 - out[..., 2]
    return out

Optional AI features

Entirely opt-in, never required for the app to work:

• Auto Narrate's AI-scripted narration path needs a local Ollama server with a model pulled (ollama pull llama3.1:8b or similar; the Setup screen shows whether it's running and links to the download) — or skip it and just type/paste your own narration text, which needs no extra setup at all.
• Auto Narrate's text-to-speech uses kokoro-tts fully offline. In the installed app the engine is built in — open ⚙ Setup → Optional features → Narration voice and click Download for the one-time ~335 MB voice models. Running from source on Python 3.13+ (kokoro-tts supports <3.13), install it separately with uv tool install kokoro-tts; the Setup screen then detects it.
• Qwikchop Deluxe can optionally hand its shortlisted highlights to a local Ollama model for smarter re-ranking — its default scoring (TextRank-style centrality, audio energy, keyword hooks) needs no network call at all.

The CLI

Everything in the GUI is a thin wrapper over the same Python CLI — use it directly for scripting, batch jobs, or just because you'd rather type a command than drag a node.

Basic usage

python3 -m videobeaux.cli --program PROGRAM --input INPUT_FILE --output OUTPUT_FILE [program options]

  -P, --program PROGRAM   Name of the effect program to run (e.g. convert, glitch)
  -i, --input INPUT       Input video file
  -o, --output OUTPUT     Output file name. No extension saves as mp4; or use .mp4/.mov/.avi/.mkv/.webm.
  -F, --force             Force overwrite output file
  -h, --help              Show help message and exit

Every program has its own flags on top of the global ones — pass -P <program> --help to see them:

python3 -m videobeaux.cli -P kinetic_captions --help

Examples

# A quick glitch pass
python3 -m videobeaux.cli -P bad_animation -i clip.mp4 -o clip_glitched.mp4

# Trim a section out of a longer video
python3 -m videobeaux.cli -P trim -i raw.mp4 -o clip.mp4 --start 1:12 --duration 8

# Burn word-pop kinetic captions from a transcript
# (transcraibe always writes <input>.json next to the input, regardless of -o)
python3 -m videobeaux.cli -P transcraibe -i clip.mp4 --stt_model models/vosk-model-en-us-0.22
python3 -m videobeaux.cli -P kinetic_captions -i clip.mp4 -o clip_captioned.mp4 --trans_json clip.json

# AI-narrated captions, fully offline (your own script, no Ollama/network involved)
python3 -m videobeaux.cli -P auto_narrate -i clip.mp4 -o clip_narrated.mp4 \
  --script "Here's what nobody tells you about this." \
  --stt_model models/vosk-model-en-us-0.22

# Pull the most interesting moments out of a longer video automatically
python3 -m videobeaux.cli -P qwikchop_deluxe -i podcast.mp4 -o highlights.mp4 \
  --stt_model models/vosk-model-en-us-0.22 --count 6 --max_time 15

Installing from source

The one thing you need first is Node.js (LTS). Everything else — Python, the video tools, all the packages — Videobeaux sets up for itself the first time it opens.

git clone <this repo>
cd videobeaux-gui/gui
npm install
npm run dev

On first launch the Setup screen gets everything ready automatically (a few minutes and an internet connection, once). If anything ever breaks, open ⚙ Setup and press Repair. The optional local-AI features (speech recognition, narration voice) are opt-in there too — they download once and then run entirely offline.

Prefer to manage Python yourself / use the CLI only?

python3 -m venv venv            # the app looks for a folder named "venv"
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install -r requirements.txt
# ffmpeg must be on your PATH. On macOS use ffmpeg-full or another libass/libzimg build
# (plain `brew install ffmpeg` lacks the `ass`, `zscale` and `drawtext` filters):
#   brew install ffmpeg-full
python3 -m videobeaux.cli --help

Building GUI installers

cd gui
npm run dist

This stages a self-contained Python environment (via python-build-standalone, with all of requirements.txt pre-installed) and static ffmpeg binaries into gui/resources/ (see gui/scripts/build-python-env.mjs), then packages everything with electron-builder into gui/dist/. Building for Windows currently needs to run on a Windows machine/CI runner (pip installs compiled wheels for whatever platform the build script itself runs on).

ffmpeg for the installer. The build bundles our ffmpeg (GPL, never --enable-nonfree). Until its release (ffmpeg-7.1.1-vb2, see .github/workflows/ffmpeg-build.yml) is published, build it locally once and npm run dist will pick it up:

cd gui
npm run ffmpeg:local        # ~15 min the first time; writes gui/.ffmpeg-build/dist/{ffmpeg,ffprobe}
npm run dist

Any folder containing ffmpeg + ffprobe also works: VIDEOBEAUX_FFMPEG_DIR=/path/to/folder npm run dist. A build configured with --enable-nonfree is refused. A locally built installer is fine for your own use; before you distribute it, publish the matching source archives (see LICENSES/README.md).

Speech-recognition models are not bundled into the installer (they run ~7GB combined) — the in-app Setup screen handles that separately, after install.

Programs

155 programs appear in the GUI sidebar (plus a few CLI-only ones), grouped the same way here. Any program can be run standalone from the CLI regardless of category.

⚡ Glitch & Corruption (21)

Signal, pixel and channel corruption — datamosh, RGB shifts, pixel sorting, warps, optical-flow smears.

Program	Description
Bad Animation	Broken-pulldown judder — wrong telecine timing makes frames stutter and comb
Bad Contrast	Harsh blend-mode contrast corruption (hard-mix, vivid-light, exclusion)
Digital Boss	Hue/saturation shift plus extreme frame-difference amplification — a busted, blown-out digital look
XRGB	Stacked red/green/blue channel displacements — heavy RGB tearing
Crossmosh	Real datamosh: decoder state corruption between two clips
Pixel Sort	Glitch-art pixel sorting — bright pixels smear into sorted streaks
VHS Tracking Error	Horizontal tracking-jitter wobble plus chroma bleed, like a worn VHS tape
Chromatic Pulse	Animated chromatic aberration that pulses over time
Deep Fry	Blown-out saturation/contrast, oversharpened, deep-fried meme look
Twociz	Extreme frame-difference amplification with a blue chroma-key knockout
Twociz Pro	Frame-difference amplification and blue chroma-key knockout, with controls
Splitting	Shuffles vertical pixel slices and blends them with the original — damaged-tape tearing
Splitting Pro	Pixel-slice shuffling with direction (vertical/horizontal/block) and slice-size controls
Slight Smear	Small red/green/blue channel offsets with wrapped edges — a subtle colour smear
Blur Pix	Pixelization with frame lag/mixing and chroma shift — smeared blocky blur
Warp	Swirl, bulge, pinch, ripple, kaleidoscope and mirror distortions, optionally animated
Flow Warp	Optical-flow smear/push — pixels drag along motion like a datamosh melt — or a color flow view
Glitch Tear	RGB channel split plus horizontal tears and scanline dimming, re-rolled every frame
Weak Signal	Edge-of-reception transmission: skew, colour misregistration, noise, streaks and bursts of static
VHS Camcorder	Home-video camcorder look: soft chroma, tape noise, a rolling tracking band and the on-screen PLAY / date stamp
CRT Monitor	Old TV tube: curved glass, scanlines, RGB shadow mask, glow and a dark vignette

≋ Trails & Echoes (15)

Feedback, ghosting and smear — effects that blend a frame with its own past.

Program	Description
Ghostee	Frame-difference amplification with colour balance — moving edges glow and ghost
LSD Feedback	Weighted multi-frame blend (including negative weights) — trippy feedback trails
LSD Feedback Pro	Multi-frame feedback blend with a configurable frame count
Double Cup	Heavy median smear blended with a weighted multi-frame mix — sludgy double image
Mirror Delay	A mirrored copy blended with a weighted multi-frame delay mix
Frame Delay Pro 1	Blends a configurable number of past frames into the output — dreamy echo trail
Frame Delay Pro 2	Lagging trail with configurable decay and YUV plane selection
Fever	Channel-plane shuffling with a frame-difference boost — feverish dream look
T-1000	Temporal median smoothing with RGB offsets — liquid-mercury shimmer on motion
Smudge	Temporal median — moving subjects smear and melt into the background
Repainting	Median repaint blended with a multi-frame mix — painterly, smeared motion
Long Exposure	A slowly fading average of past frames — moving things smear into ghostly trails
RGB Time Split	Red channel is now, green and blue lag behind — motion leaves rainbow fringes
Feedback Loop	Video feedback: the output is fed back zoomed, rotated, shifted and hue-shifted each frame so the picture spirals into itself
Key Feedback	Video feedback with two keys — a luma/color Insert key picks what enters the loop, a Feedback key picks what survives each pass — plus zoom, rotate, shift and hue shift

⏱ Time & Motion (18)

Speed, reversal, loops, stutters, freezes and scrolling.

Program	Description
Speed	Change playback speed without pitch-shifting audio
Reverse	Reverse the video
Boomerang	Forward-then-reverse ping-pong loop
Time Ramp	Variable speed ramp within one clip (e.g. slow-mo into a speed-up), unlike the flat Speed effect
Freeze Punch	Freezes on detected audio peaks then resumes — a punchy freeze-frame emphasis edit
Strobe Cut	Periodic flash/strobe brightness spikes at a configurable interval
Stutter Pro	Replaces frames with random picks from the last N frames
Nostalgic Stutter	Random-frame stutter with chroma shift and multi-frame mixing, like a corrupted file
Overexposed Stutter	Hard blend modes with random-frame repeats and lag — blown-out corrupted stutter
Looper Pro	Repeats a chosen segment (start frame + length) a set number of times
Frame Interpolate	Paints in-between frames: smoother motion (30→60 fps) or smooth slow motion. Slow on long clips
Scrolling Pro	Scroll the picture horizontally and/or vertically at a set speed
Broken Scroll	Amplified frame differences plus a slow vertical scroll — rolling broken-tracking look
Slit-Scan	Different parts of the frame show different moments in time — rows, rings, columns or waves of delay
Tunnel	The picture wrapped around the inside of a tunnel you fly down
Little Planet	Polar-coordinate 'tiny planet' — the bottom of the picture becomes a small round world with sky all around
Strobe Hold	Stroboscope: hold each picture for N frames (stuttering low-frame-rate look), with optional blink color and random holds
Freeze Frame	Freeze the picture at a chosen moment — hold in place (same length) or insert the still (longer clip, silence under it)

✦ Color & Look (49)

Colour grading and stylization — grades, LUTs, film/night-vision looks, halftone, duotone, dithering, cartoon and sketch.

Program	Description
Gamma Fix	Adjust gamma, brightness, contrast, and saturation
LUT Apply	Apply a 3D LUT (.cube / .3dl) with optional colour adjustments
Duotone	Maps luminance to a 2-color gradient — a stylized-poster look
Night Vision	Green-phosphor night-vision-goggle look with grain and vignette
Old Film Damage	Vintage film-print scratches, dust, flicker, and gate-weave
Halftone	Newsprint-style halftone dot pattern, dot size driven by brightness
Steel Wash	Shear plus a cold steel-blue vibrance grade
Pickle Juice	Shear plus a strongly green-skewed vibrance grade
Septic	Green/magenta-skewed vibrance grade — a sickly colour cast
WB Flare	Wide bilateral blur — soft, blown-out white-balance glow
WB Flare Pro	Bilateral blur with a configurable sigma
Xpiritualism	Multi-layer bloom with a pastel colour pass — soft, dreamy glow
Zapruder	Frame-difference amplification with colour correction and DCT denoise — degraded found-footage look
Bad Predator	Heat-vision look — amplified frame differences with a hot false-colour grade
Ball Point Pen	Frame-difference edges with deinterlace artifacts and colour balance — inked sketch look
Light Snow	Motion-interpolated frames with chroma shift and debanding — a light static shimmer
RB Blur	Debanding (gradfun) — smooths banding in flat gradients like skies and shadows
RB Blur Pro	Debanding (gradfun) with strength and radius controls
Recalled Sensor	Strong frame-difference bloom — edges burn bright like overexposure
Recalled Sensor Pro	Frame-difference bloom with radius and intensity controls
Soapblind	Heavy wavelet denoise — plasticky, smoothed, soap-in-the-eyes look
Dither	Adaptive-palette dithering — Floyd-Steinberg, Atkinson, Sierra, Bayer and more, with chunky Pixel Size
Retro Dither	Dither onto a classic fixed palette — 1-bit B&W, Game Boy, CGA, EGA, C64, PICO-8, phosphors, or your own colors
Ordered Dither	Bayer, clustered-dot, blue-noise and static dither patterns onto a palette or posterized colors, optionally animated
Neon Edges	Glowing colored edge outlines over a dimmed, original or black background
Cartoon	Flat posterized colors with inked outlines
Sketch	Pencil, colored-pencil, watercolor-style and painterly looks
K-Means Palette	Snap the video to its N dominant colors, optionally dithered
Filter Library	Every photo-booth filter in one list, plus any you drop into ~/.videobeaux/filters/ (Game Boy, CGA/C64/PICO-8, halftone, comic, fisheye… also live in Retro Dither, Halftone, Cartoon, Warp)
ASCII Art	Rebuild the video from text characters — choose character set, colors (matrix green, amber…), size and edge boost
Negative	Invert the picture like a photo negative — or flip only the brightness and keep the colors
Black & White	High-contrast black and white with local contrast boost (faces and texture pop), optional film grain
Sepia & Tones	Antique single-tone looks: sepia, cyanotype blue, rose, forest or gold
Thermal	Thermal-camera false color — pick the heat palette (inferno, jet, turbo, hot, plasma…)
Infrared Film	False-color infrared film look — foliage turns pink and red, skies go dark
Solarize	Darkroom solarization — tones above the threshold flip, giving glowing metallic edges
Posterize	Reduce each color channel to a few flat levels
Lomo	Cross-processed toy-camera look: punchy curves, color cast and dark vignette corners
Hue Cycle	Rotate every color around the color wheel continuously
Pop Art	Warhol-style flat-color panels — four colorways in a 2×2 grid, or one palette over the whole frame
Blueprint	Technical-drawing look — white edge lines on blueprint blue, with an optional grid
Emboss	Raised-relief emboss lit from any angle, in gray or keeping the colors
Oil Paint	Painterly oil-paint look: smoothed brush regions, posterized tones and a little canvas relief
LED Wall	The picture rebuilt from a grid of round glowing LEDs, like a stadium video wall
Pixelate	Chunky square pixels — a mosaic of any block size
Color Pass	Keep one color range and turn everything else gray (or remove just that color) — a red dress in a gray world
Proc Amp	Video processing amp: brightness, contrast, saturation, hue rotation, gamma, black/white levels, color temperature, broadcast-safe clamp
Beauxtrix	A video blending matrix (homage to the LZX Video Blending Matrix): three R/G/B mixers summing videos A–D with −2…+2 levels, bias, and sum or absolute (solarize) outputs
Topobeaux	Slice the picture into up to 6 tonal bands (luma, color channel, saturation, hue or motion) — false color, tint, contour lines, a different video per band, or a matte cutout. Homage to the LZX Topogram

◉ Vision & Tracking (7)

Computer vision (OpenCV) — face tracking, redaction and reframing, motion isolation, feature tracking.

Program	Description
Face Track	Detect and track faces with boxes, brackets, IDs, trails, a spotlight, or an image pasted on each face
Face Redact	Blur, pixelate, fill or dither over tracked faces — or hide everything except the faces
Face Follow	Smart reframe: a smoothed virtual camera that pans and zooms to keep a face in shot
Motion Ghost	Isolate what moves — tint it, show only the movers, or leave glowing motion trails
Feature Trails	Tracking-HUD look: tracked points, trails and connecting lines over the video
Face Warp	Big head, tiny head or big eyes — warps tracked faces (works on several faces at once)
Face Swap	Swap the two biggest faces in the shot, colour-matched with a soft edge

⊞ Layout & Overlay (13)

Put things on top of or next to each other — overlays, stacks, multi-layer composites.

Program	Description
Watermark / Image Overlay	Overlay a watermark or image onto the video — 9-point placement or custom X/Y, scale or exact pixel sizing, opacity, spin, and a timed enable window
Remove Background	Cut the subject out and put anything behind it — or export transparent WebM/MOV. Static-camera mode needs no download; optional local AI models (U²-Net) for harder shots
Chroma Key	Remove a green/blue screen (or any solid color) and put a color, image or another video behind — or export transparent WebM/MOV. Auto-detects the screen color
Luma Key	Knock out the darks or brights (black backgrounds, white skies); screen/add blend modes for fire, smoke and light leaks
Stack 2×	Stack two videos vertically (input on top, input2 on bottom)
Triptych	Arrange three videos in a symmetric hstack or vstack layout
Lagkage	JSON-driven multilayer compositor
Layer Blend	Layer two videos with per-layer opacity and a blend mode — multiply, screen, color burn, difference, overlay and more. Choose whose audio to keep
Picture-in-Picture	A small second video inset over the main one — pick the corner, size, border, opacity and whose audio you hear; swap to flip which is full-screen
Quad Split	Split screen: up to four videos in a 2×2 grid, side by side, stacked, or one big plus three small, with adjustable gaps
Video Wall	The picture repeated in a grid of tiles — plain repeats, mirrored tiles, or a delay wall where each tile lags a bit more
Rotate	Turn the picture upside down, on its side, tilted at any angle (45°…) or slowly spinning — the output file really is rotated; choose to grow the canvas, zoom to fill, or keep the frame size
Shape Synth	Two animated shape generators (circle, ring, diamond, square, cross, stripes, checker, vortex) with folding, curves, stretch, spin and drift — painted over, used as a window, a warp, or a cutout. Homage to the LZX DSG3
Stackadelic	Three stacked rectangle / rhombus layers with layer priority — each cuts out of the ones beneath, showing a color, input A/B/C or A keyed by brightness; tile into mosaics. Homage to the LZX Stacker

✂ Cut & Assemble (8)

Cut videos apart and assemble them — trims, splits, joins, inserts, transitions.

Program	Description
Trim	Extracts a single section of a video by timestamp — grab a clip from the middle, or trim off the start.
Qwikchop	Split a video into exactly N equal segments, exported as separate files. Optional seamless-head trimming to avoid black flashes at cuts.
Mince	Merge a folder of videos into one output in a chosen order
Concat	Join two videos back to back — first then second. Good for adding a slate before the main video.
Insert Clip	Inserts a second video into the master at a chosen timestamp, then picks up the master from where it left off — e.g. an intermission slate. Independent transition control at each boundary.
Wipe Transitions	Combine two videos with a transitional wipe using ffmpeg's xfade filter
Fade & Flash	Fade in/out from black, white or any color, plus timed flashes that decay or snap on/off; optionally fades the audio too
Add Audio	Put your own music or audio under a video — replace the sound, mix it in, or duck it under the original voice; picture is copied untouched

◈ Speech & Captions (9)

Transcript-driven editing, captions and narration (Vosk / kokoro-tts, all local).

Program	Description
Transcraibe	AI speech-to-text transcription (Vosk)
Kinetic Captions	Per-frame rendered captions with a true size pop on the active word
Captburn	Burn subtitles / captions into video
Auto Narrate	Adds AI-narrated captions to a video — type your own script (offline, default) or draft one from a topic via an optional local Ollama model. Synthesizes speech (kokoro-tts), transcribes it locally (Vosk), mixes it over the original audio, and burns kinetic-typography captions (the spoken word pops larger, per-frame rendered — see Kinetic Captions).
Grep Supercut	Searches a transcript for a word, phrase, or regex and cuts together every match — the classic "videogrep": find every time someone says X.
Qwikchop Deluxe	Content-aware highlight extraction: scores candidate excerpts from the transcript (local centrality + audio energy + keyword hooks — no network) and exports the best ones, each as its own file.
Silence Xtraction	Remove speech segments, keeping what's left (not exactly silence, but no discernable words)
Word Xtraction VAD	Remove recognized speech (Vosk) and optionally local Silero VAD detections, keeping applause, laughter, music, noise, and other non-speech audio.
Ngrams	Lists the most common word sequences spoken in a transcript — useful for finding good search terms before running Grep Supercut.

⚙ Media Tools (14)

Convert, resize, stabilize, extract, inspect and download.

Program	Description
Convert	General-purpose converter — codec, quality, and format control. Output container is set by the Output node.
Convert Dims	Convert and change video dimensions
Resize	Resize video to specific dimensions
Shortify	Converts a video to vertical Shorts/Reels format with a blurred, enlarged background fill instead of black bars.
Tonemap HDR→SDR	Tonemap HDR content to SDR
Extract Frames	Extract frames from video as images
Extract Sound	Extract audio track from video
Thumbs	Generate thumbnail grid from video
Subs Convert	Convert subtitle format
Download YT	Download video from YouTube / yt-dlp
Media Info	Concise report of dimensions, duration, fps, codecs and size — printed to the log and saved as JSON
Meta Extraction	Extract video metadata / ffprobe info
Hash Fingerprint	Generate a perceptual hash fingerprint
Stabilize	Remove camera shake with feature tracking and a smoothed camera path

_{Three more programs exist for backward compatibility, CLI-only (not in the GUI's sidebar): convert (superseded by Convert above), chain_builder, chain_builder_pro.}