[CoffeeFlux]

The Misc. APIs page describes aegisub.frame_from_ms and aegisub.ms_from_frame as exact frame-boundary conversions, but they don't work that way. In src/auto4_lua.cpp both call the video controller with agi::vfr::START, so:

• frame_from_ms(ms) returns the first frame that begins at or after ms, which is the first frame a line starting at ms is shown on. A time in the middle of a frame rounds up, not down as the docs say.
• ms_from_frame(frame) returns prev + (cur - prev + 1) / 2 (Framerate::TimeAtFrame in libaegisub/common/vfr.cpp), the midpoint between the previous frame's start and this frame's start. It is not the first millisecond inside the frame. At 23.976 fps, frame 24 begins at 1001 ms and ms_from_frame(24) returns 980.

This has been the case since at least the 2019 Aegisub/Aegisub master, so this updates both the latest and the 3.5 docs. It also adds a short paragraph explaining that these are subtitle times, and why the midpoint is useful (it survives centisecond rounding and round-trips with frame_from_ms).

I checked the claims against a Python port of FrameAtTime/TimeAtFrame:

• Round-tripping and surviving centisecond rounding hold at 23.976, 25, 29.97, 50, 59.94, 60 and 100 fps, and for a mixed VFR timeline.
• Centisecond rounding stops working above 100 fps, which the docs now mention.

The zh-cn translation and the old 3.0–3.2 pages aren't changed.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
aegisite-nycd	Ready	Preview	Oct 25, 2026 10:35pm UTC

[arch1t3cht]

Thanks!

I find the "A time at which a line starting there is first displayed on frame" part a bit confusing. Maybe this is better:

The start time that a subtitle line should have if it is supposed to first become visible at the frame frame. This is also the end time that a subtitle line should have if it is supposed to first become invisible at the frame frame. If no frame rate data is loaded, returns nil.

The rest LGTM.