Skip to content

QuickEdit v0.4.0

Choose a tag to compare

@Zovelik Zovelik released this 04 Nov 22:08
· 5 commits to master since this release

Shaders that listen better and remember more: generators can now draw in several passes with feedback from the frame before, and read the sound of every track in stereo. QuickEdit also has a website with a guide.

Shaders

Multi-pass generators

  • A generator can define bufferA, bufferB and bufferC (vec4 bufferA(vec2 uv, float time)). They are drawn first, in that order, into half float pictures the size of the frame, and kept by clip.
  • passA(uv), passB(uv), passC(uv) read the passes drawn before, for this frame.
  • lastA(uv), lastB(uv), lastC(uv) read the passes of the frame before, for feedback effects such as trails, smoke or echoes. They are cleared to black after a jump in time.

Sound in stereo, track by track

  • spectrum(track, channel, x): the spectrum of the last 85 ms, x from 20 Hz to 20 kHz on a log scale, 0 to 1 for -90 to -10 dB.
  • waveform(track, channel, x): the last 512 samples, -1 to 1.
  • Track 0 is the mix, then the sound tracks S1, S2..., then the audio tracks A1, A2...; soundTracks() gives how many there are. Channel 0 is left, 1 is right. Muted tracks are silent.
  • Works the same in the preview and in the export.

Sound history

  • soundBefore(seconds): the sound level up to 1.5 s before the frame, for pictures that remember the beat.

The generator template in the shader editor documents all of the above.

MCP server

  • list_capture_sources lists the screens and windows again each time it is called, so windows opened after the first listing show up.
  • write_shader describes the new shader functions (multi-pass, spectrum, waveform, soundBefore), so AI assistants can write these shaders.

Website

  • A landing page and a guide, published on GitHub Pages.

Full changelog: v0.3.0...v0.4.0