You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
HTML: frame 0 is drawn before the GSAP timeline's time-0 sets apply (wrong poster frame) #79
In an HTML (HyperFrames-style) composition, frame 0 is drawn without the GSAP timeline's state at time 0. tl.set(…, 0) calls are not applied to the first frame, so it shows the page's raw CSS. Frame 1 onward is correct.
This affects both raea render PAGE.html -o OUT.mp4 (the mp4's first frame) and --at 0. For a composition whose frame 0 is its poster/thumbnail, the poster comes out wrong: in our case every caption that a set hides at 0 is drawn on top of each other.
Repro (raea 0.1.0, linux-x86_64 release)
index.html, with GSAP 3.14.2's gsap.min.js next to it:
Expected (as in Chrome / HyperFrames), frames 0 and 1 identical: only B, shifted 200 px right. Actual:
frame-00000.png: A drawn in red at full opacity, B unshifted, both texts overlapping at x = 40.
frame-00001.png: correct (A hidden, B at x = 240).
Environment
raea 0.1.0 (release tarball, sha256 checked), Linux x86_64
Reproduced on Ubuntu 26.04 with an Intel HD 630 (Mesa Vulkan), and on Ubuntu 24.04 under WSL2 with llvmpipe, so it doesn't depend on the GPU.
Found while rendering a 1,800-frame GSAP showcase: frames 1–1799 matched the HyperFrames render, and only frame 0 differed.
Possibly related
The same page logs usvg … Failed to parse transform value: 'translate(1500px, 1000px)' for an inline <svg> that GSAP moves with x/y. GSAP writes px units, which SVG's transform attribute grammar doesn't allow. The element still ends up in the right place on later frames, so this is just noise in the log, but it may be worth handling the same way #60 handled stylesheet transforms.
I hit this too and traced it: a paused GSAP timeline's playhead starts at 0, so totalTime(0) is a no-op and the time-0 sets never render. Reproduces in plain GSAP 3.14.2, no raea needed.
A 2-line fix in sdk/ui/src/html/page/time.tsseek(): at t=0, call tl.render(0, false, true) instead. I have it with a regression test (fails without it, all sdk/ui tests pass with it). Happy to open a PR if you'd like one.
Activity
Ragulsundaram commentedon Nov 10, 2026
I hit this too and traced it: a paused GSAP timeline's playhead starts at 0, so
totalTime(0)is a no-op and the time-0 sets never render. Reproduces in plain GSAP 3.14.2, no raea needed.A 2-line fix in
sdk/ui/src/html/page/time.tsseek(): at t=0, calltl.render(0, false, true)instead. I have it with a regression test (fails without it, all sdk/ui tests pass with it). Happy to open a PR if you'd like one.HTML: frame 0 applies the GSAP timeline's time-0 sets (#79)
Merge public PR #81 from Ragulsundaram: HTML: frame 0 applies the GSA…
shinyobjectz commentedon Nov 10, 2026
Fixed by #81 (merged as 4eb42e3): at t=0 the seek forces the timeline's render, so frame 0 has its time-0 sets. It will be in the next release.
skills pin: raea-skills #79 (9bdd33c): reference/files.md names the r…