Burn Subtitles into Video

• Choose one local video and a UTF-8 SRT or WebVTT caption file.
• Set caption size and margin so the text fits the video. Choose WebM or MP4 output.
• Process video, review the reported cue count and any clipped or omitted times, then download the video. Captions become permanent pixels; the original soundtrack is retained.

A caption from 00:00:00,250 to 00:00:00,750 appears for exactly that half-second, even if the source frame holds across the cue boundary.

• 128 MiB total local input/output; 5 minutes output; 8 megapixels/frame and 8192 pixels/side. Video edits are bounded to 9,000 frames and 4 billion decoded/output pixels including backward seeking, with a 120-second cancellable worker deadline.
• Local MP4/MOV/WebM/Matroska with browser-supported codecs. Primary tracks only. WebM uses VP8/Opus; MP4 uses AVC/AAC when supported by the current browser. No network source or server fallback. Output is re-encoded; separate alpha-plane video is unsupported, supported transparency becomes black, and source tags are omitted, and HDR/color-profile fidelity is not guaranteed.
• Video join fits clips into one even-sized canvas with black letterboxing. Container display orientation is applied. Speed changes preserve synchronization by changing audio speed and pitch together. Reverse is limited to 60 seconds and 600 frames; backwards decoding can be slow.
• Decoded audio is limited to 5 minutes total, 32 million samples and 1–2 channels. Audio is processed at 48 kHz. Added sound is trimmed or padded/looped to video length; mixed sound uses peak protection. Gain above full scale may clip.
• PNG sequence export is limited to 300 images and 100 million output pixels and includes requested/source timestamps. Animated WebP uses lossy browser encoding, no audio, millisecond-rounded delays, 300 frames, 20 million total frame pixels and 40 MiB output.
• Image input is limited to 30 static PNG/JPEG/WebP images, 20 MiB each and 128 MiB combined. Stored pixel orientation is used. Text uses the local sans-serif font and is limited to 200 characters/five lines; overlays must fit within the output frame.
• Subtitle input: UTF-8 up to 1 MiB, 2,000 cues, 500 characters/six explicit lines per cue, eight simultaneous cues and 12 wrapped lines total. Plain white local-font text on a translucent black box; formatting, WebVTT style/region/position settings and karaoke timestamps are unsupported. Source tags are omitted.
• Cue times are clipped to the video duration and outside-video cues are omitted with counts. Overlapping cues stack in time order. Added cue boundaries count against 9,000 output frames/four billion pixel work. WebM frame/cue timing rounds to milliseconds and coincident boundaries are combined. Captions that do not fit vertically fail with an actionable error.
• When audio is decoded, tracks with more than one second of decoder priming or overlapping sample timestamps are unsupported. The 32-million decoded-sample work cap includes priming and skipped samples.