New 'Area LED' controls in the Schermo tab: W x H @ X,Y. Everything
(canvas, videos, images, logos, ticker) lives in a #stage box whose
transform makes it the containing block for the fixed layers, so resizing
it confines the whole show at once, black outside. The canvas resizes to
the stage, so shaders render at the wall's native resolution. Persisted
in localStorage and re-pushed when the output window (re)starts; the
recorder follows the canvas position.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
recStart now carries the format and an optional maxMs; the output stops
itself when the time is up (15/30/60/90s picker in the control UI), so a
reel doesn't depend on a manual stop.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three modes for where the director draws presets from: 'smart' keeps the
curated per-mood pools, 'all' removes the family restriction, 'custom'
limits picks to a user checklist (mood still shapes the speed variant,
with a fallback when the checklist exhausts it). Config persists in
localStorage and is re-pushed live while the director runs.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The audio engine now tracks musical features: fast/slow energy EMAs, BPM
(median of beat intervals), a quiet/groove/peak mood with hysteresis, and
drop detection (fast energy diverging from slow right after a lull). An
output-side director, toggled from the control UI, switches presets every
16 beats (or ~14s without beats) and immediately on a drop, drawing from
mood-matched family pools (scenic for quiet, fluid/geometry for groove,
fractals/tunnels/fire for peak) while avoiding recent repeats. A manual
effect pick takes over and turns the director off.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The packaged app doesn't bundle ffmpeg, so on Macs without a system ffmpeg the
waveform stayed blank. Compute peaks by reading the file bytes (new file:read
IPC) and decoding with Web Audio's decodeAudioData in a low-rate AudioContext;
ffmpeg in main remains only as a fallback.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Pointer on the waveform now seeks: scrub the playhead on the playing track
(new 'seek' output message setting currentTime) or start a stopped track from
the clicked point (playIndex(i, fromTime)). Grabbing within ~10px of the green/
red handles still adjusts the trim in/out instead of seeking.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Per-track start/end points (mm:ss inputs, 'set to current' buttons, draggable
handles on the waveform). Trim applied via audio.setTrim (seek to start, stop
at end) and pushed live to the playing track.
- Wider cards showing elapsed / total / remaining for the trimmed range.
- Waveform rendered per track: peaks computed in main via ffmpeg (mono 8kHz PCM
-> 400 normalized buckets, cached), drawn on a canvas with trim region +
playhead. New audio:peaks IPC + djv.peaks preload binding.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Add interlude playlist entries with no audio that run their own scene for a set
duration, then auto-advance. Shares a unified timed-segment runner with the
inter-track gap (pause/resume via the Play button). New 'playSilence' output
message + audio.silence() to stop audio while keeping meters/recording alive.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- When a playlist video has an active scene effect, auto-apply 'screen' blend so
the effect shows through (only if the user left blend at default 'normal';
manual choices are respected). Restores on effect-off / next track.
- Per-track 'Pausa dopo' (seconds): after a track ends, keep its scene timeline
running for N seconds as an intermission (scene elements placed past the song
length appear during the pause), then advance. Countdown + badge in the row.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
On restart, lastScene was reset to null and a channel with nothing active at
t=0 also evaluated to null, so the diff saw 'no change' and never sent the hide
command — leftover scene text stayed on screen. Use an undefined sentinel so
advanceCues applies (and thus clears) every channel once at scene start.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Each scene element (effect, text, image) now has its own appearance time and
optional duration, with all styling options inline (no more capture-from-panel).
Live preview applies edits immediately on the playing track. Moved the playlist
and its video-render controls into a new Playlist tab.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Blend mode, opacity and fit for videos played from the playlist (mix with visuals).
- Save/Load buttons to export/import a playlist (with scenes/cues) to a JSON file
via native dialogs; the playlist already auto-persists between sessions.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Audio and video files can both be added to the playlist; video tracks play
full-screen with their audio routed through the analyser (visuals/waveform
react), have duration/progress, and are included in recordings. Persisted.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Loads a muted looping video as a layer above the WebGL canvas with opacity,
blend mode and fit controls; included in MP4 recordings and droppable.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Each playlist track holds a timeline of cues (@time → effect/text/image);
cues fire automatically as the track plays, driven by progress reports.
- Playlist (tracks, hotkeys, cues) is saved to userData and restored on startup,
like the pad bank.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Each playlist track gets an optional scene (🎬) bound to it: pick the current
effect, a ticker text and an overlay image. When the track plays, the scene is
applied automatically — a cue/queue per song. Scene image is a dedicated output
layer (also captured in recordings).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Scrolling text: font, size, weight, colour and movement effects (up/down, wave,
zoom, flash, rotate) rendered per-letter (and captured in recordings).
- Larger multiline text input.
- Pad bank: crossfade between pads (per-source gain ramps in the audio engine).
- Duration + remaining-time and progress bars for the playlist and the active pad,
via metadata probing and live progress reports from the output window.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Electron app with audio-reactive fractal/abstract visuals on an external
display. Dual-window setup (control panel + fullscreen output), 8 WebGL
scenes, audio from file/live input/system audio, image and scrolling-text
overlays.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
