Feature Library

Rivers of glyphs

3.63 GPU · 0.13 CPU ms
A still of the Rivers of glyphs demo
Still
The live demo starts when it’s in view.Drag or use the arrow keys to look around · Space plays or pauses · R replays

A field of characters wakes: every glyph lifts off its cell, rises on a curl flow into a few rivers of light, runs along them and pours into a word, where one…

What it costs a frame

3.63 ms
0.13 ms
GPU
3.63 ms
CPU, median
0.13 ms
CPU, 95th pct
0.20 ms
Draw calls
0
Weight
Heavy

Measured 11 November 2026 on 13th Gen Intel(R) Core(TM) i9-13900H, Chrome 157, 1920×1080 at 1×. 600 frames × 3 rounds, median kept. What ran: Measured looping by itself with about 130,000 glyphs: the field wakes, the rivers run and pour into the word over the play time, it holds a second and replays. Every glyph is computed every frame (stateless); one pass draws the field per pixel and one the glyphs, then a quarter-size bloom. Timed with its own timestamp queries. Phones: tiers are being measured next.

On every machine it was measured on
MachineGPU msCPU msCPU p95
unknown GPU · 13th Gen Intel(R) Core(TM) i9-13900Hreference · measuring tool3.630.130.20
apple · Apple A19 Proin a browser tab · 16 November 20266.710.280.34
Intel Arc 140V (Xe2-LPG) · Intel Core Ultra 7 258Vin a browser tab · 15 November 20261.440.300.53

Bars against one 16.7 ms frame. A browser-tab run is timed in a visible tab without the tool’s scripted input, so read it as close, not exact.

What it needs

Renders with
WebGPUNeeds WebGPU: current Chrome and Edge, Safari 26+, Firefox 141+ on Windows.
Your browser
Checking…Looking for WebGPU
Depends on
NothingNo other packages

What it’s for

  1. Title and splash moments
  2. Data or AI themed openers
  3. A word or a number assembling on a stage
  4. Transitions between sections
  5. Loading moments that end on a result
  6. Ambient data backdrops

Pitfalls, in brief

  1. Stateless means every glyph is computed every frame, waiting or not: the cost is the field’s size, not how many are moving.
  2. Every point of the word keeps exactly one glyph only when the field has more glyphs than the word has points: a big word on a small field leaves holes.
  3. Progress must reach exactly 1 for the word to be held still: clamp it, and use heldFrame() to skip draws while nothing can change.
  4. The streaks need to know how fast progress moves (velocity) and how fast time runs (timeScale): pass 0 when scrubbing, or a drag smears every glyph.
  5. Load the font before drawing the word mask and before creating the rivers (the glyph atlas is drawn then): a fallback font is baked in otherwise.
3 more in brief
  1. Never queue more than one frame ahead: draw() skips while its last frame is still on the GPU.
  2. Changing the field (its size, seed or origin) means building the field and the rivers again; a new word or place only re-assigns the glyphs (setWord), no pipelines rebuilt.
  3. Without WebGPU, draw the still (drawRiversStill): the word set in its glyphs over the dark field, the same picture the rivers end on.
8 pitfalls in full, the hook-up guide, the module source and its tests come with the part.

Where it comes from

From SpaciousPro. Fully custom code.

Mark Jarzynski and Marc Olano, “Hash Functions for GPU Rendering” (JCGT 2020): The PCG hash behind every glyph’s randomness, in TypeScript and WGSL alike.

Robert Bridson, Jim Hourihan and Marcus Nordenstam, “Curl-Noise for Procedural Fluid Flow” (SIGGRAPH 2007): The idea behind the flow: the curl of a potential is divergence-free, so the glyphs swirl without bunching. Here the potential is a few drifting waves, not noise.

computeraw webgpuwgslinstancingcurl noiseglyph atlastext effectsstatelessscrubbabledata streamsbloom

Near it in the frame