Discord per-file upload limits, the video bitrate they leave, and two measured browser encodes

Measured data about Discord's per-file upload limits and the video bitrate those limits leave you, plus the raw ffprobe output for two encodes that were actually run. What is in here discord-file-size-limits.csv - the per-file upload ceiling for each Discord account tier, in bytes, with the official page each figure came from, the sentence quoted from that page, and the date it was checked. Where two official Discord pages disagree above 50 MB, both figures are recorded with their sources rather than one being picked. bitrate-budget.csv - for an 8, 10, 20 or 50 MB ceiling and a clip of 15 s to 3 min, the video bitrate that fits, after a 0.6% container reserve and the audio track are subtracted. measured-encodes.csv - two encodes run end to end: what was asked for, what came out, the input file's SHA-256, the encoder settings, and the wall-clock seconds. quality-levers.csv - what changing resolution, frame rate, x264 preset and audio bitrate each costs or buys, measured, with the record file each row comes from. site-10mb-10s-reserve.probe.json, site-3mb-10s-audio-reserve.probe.json - the unedited ffprobe output for the two encodes above. How it was measured Both encodes were run in a browser tab, through FFmpeg compiled to WebAssembly (ffmpeg-core.wasm), with libx264 at preset veryfast, video in yuv420p, AAC audio, and the MP4 written with +faststart. No server was involved and the input files never left the machine. Run 1: input 15,942,580 bytes, 1280x720 at 30 fps, no audio track; target 10 MB; 8,274 kbps requested on the first pass, which overshot, so a second pass was run at 7,032 kbps; output 9,253,520 bytes (video measured at 7,399,220 bps) in 140 s. Run 2: input 5,407,489 bytes, 1920x1080 at 60 fps with stereo AAC audio; target 3 MB; 2,437 kbps requested, one pass; output 3,138,651 bytes (video 2,432,889 bps, audio 63,940 bps) in 150 s. Conditions and limits Elapsed seconds are wall-clock time on the single machine that ran the tests and are not a hardware benchmark. The bitrate figures are what the encoder reported and what the output file measured, not theoretical targets. Discord's own limits were read from Discord's published support pages on 2026-11-02; the limits can change, and the CSV records the exact sentence each number came from so a later reader can tell whether the page still says the same thing. The tool these numbers were measured on is at compressmyclip.com - same encoder, same container reserve, nothing uploaded.