agentleFS
JagZ999JagZ999/explainer-video/skills/smooth-render/SKILL.md

smooth-render

Skill1 starsChanged 4 months agoMITView on GitHub

Render an HTML/CSS animation to a smooth, sharp MP4 at native resolution (up to 2K/4K) without choppy motion. Use when a CSS/JS animation (especially glassy/blur/backdrop-filter, gradient, or continuous-motion scenes) renders JANKY or stuttery via screenshot/screencast capture, when you need per-frame-accurate logical-time output with no wall-clock drift, or when muxing audio onto a rendered animation. Triggers: "render my animation to video", "the capture is choppy/stuttering", "export this HTML animation as MP4", "2K render of my web animation", "frames are dropping when I record the page".

The file

Files
SKILL.md

Smooth Render

Capture an HTML/CSS animation to a crisp MP4 where motion is buttery, not choppy — even for the cases that normally stutter: backdrop-filter/blur, glassy cards, gradients, and always-on continuous animations.

Why naive capture is choppy

Chrome's CDP screencast (and most "screenshot every frame" loops) coalesces gradual changes — a smooth CSS transition emits far fewer frames than it should, and the result judders. Slowing the output fps doesn't fix the missing frames.

The fix (two tricks, in scripts/render.js)

  1. Slow logical clock — the page advances its own time t by dt * window.__RATE (e.g. RATE=0.7). Each logical second now spans more wall-clock time, so the screencast emits more frames per logical second. Every captured frame is tagged with the page's logical t, then remapped on the way out → exact timing, no drift.
  2. Heartbeat — a 1px corner dot mutated every requestAnimationFrame so the screencast never sees a "still" region and keeps emitting at a uniform high rate.

Output is per-frame JPEGs + frames.json (per-frame logical timestamps). scripts/assemble.js turns that into an ffmpeg concat list with per-frame durations (= logical-time diffs), so the final encode is perfectly timed regardless of capture jitter.

Wiring your page (required)

The page must expose globals t (logical seconds), TOTAL, seek(t), play(), and in its rAF loop advance time by the rate:

function tick(now){ if(!playing) return;
  if (last==null) last=now; t += ((now-last)/1000) * (window.__RATE||1); last=now;  // slow clock
  if (t>=TOTAL) { t=TOTAL; pause(); }
  render(t); requestAnimationFrame(tick);
}

Scale your CSS durations by 1/RATE so motion stays correctly timed in logical time. Either author them as calc(var(--k)*…) and set --k from __TK, or inject a stylesheet at render time that multiplies every transition-duration/animation-duration/delay you use by K = 1/RATE. Keep one scaled rule per animated selector — whenever you add or change a duration in the page, update its scaled rule too (the #1 source of "render timing looks off").

Usage

# serve the page (a local static server avoids file:// quirks)
python3 -m http.server 8755 &
RENDER_URL=http://localhost:8755/index.html RENDER_W=2560 RENDER_H=1440 RENDER_RATE=0.7 \
  NODE_PATH=$(npm root) node scripts/render.js          # -> frames/ + frames/frames.json
node scripts/assemble.js frames                          # -> frames/list.txt
( cd frames && ffmpeg -y -f concat -safe 0 -i list.txt -i ../audio.mp3 \
    -vf "fps=30,format=yuv420p" -c:v libx264 -crf 18 -preset medium \
    -c:a aac -b:a 192k -movflags +faststart -shortest ../out.mp4 )

Drop the -i ../audio.mp3 / audio flags for a silent render.

Knobs & gotchas

  • RATE: 0.6–0.8 is the sweet spot (still ≥30 logical fps, far fewer frames than 0.28). Lower RATE = more frames + bigger frame cache + slower. Pages with lots of always-on animation saturate the screencast quickly → prefer the higher end.
  • GPU flags (--use-angle=metal, --enable-gpu-rasterization) keep 2K capture from being throttled by software compositing. Drop backdrop-filter (blur) at capture time — it's per-frame expensive and visually negligible over a soft gradient.
  • Disk: frames are small (~0.1 MB at 2K), but a 2–3 min video is thousands of files — delete frames/ after encoding.
  • Audio: render silent, then mux a separately-built track onto the logical timeline (see the audio-stems skill). Mute/stub any <audio> during capture.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.