Document .opus emote sound and add a converter for it

[dima-dencep]

Emote sound moved from Note Block Studio to Opus, so the page that told people to open Note Block Studio needed rewriting, and the format it now describes is one nobody can produce by hand without reading a spec.

creatingmusic.md

Rewritten around putting an .opus next to the emote:

• what .opus even is — ordinary audio, the same format Discord uses, chosen here because it stays small enough to travel inside the emote packet — and how to play one back: Windows, Linux and Android open it as is, macOS needs VLC, and dragging the file into a browser works anywhere
• where the file goes and that the name has to match the emote
• the LOOPSTART tag, with the ffmpeg command and the note that 48000 samples is one second
• the limits (mono, 1 MB, 96 kbps, 10 minutes) and which one you actually run into — at 32 kbps that is about four and a half minutes
• why you should not normalise the file yourself: the mod evens out loudness between emotes, and players can turn that off
• Note Block Studio kept at the end as legacy. Current versions no longer play .nbs, but it is still passed through to older clients, so keeping both files next to the emote is worth explaining

/opus-converter

Making that file by hand means knowing the mod wants mono at 48 kHz and that the result has to survive a network packet. This page does it instead: drop in any audio file, choose quality, trim and looping, download something the mod accepts.

• built as a Docusaurus page, so it gets the site theme, dark mode and navigation for free
• mono and 48 kHz are applied without asking
• the size is estimated before encoding, shown against the 1 MB limit, and the button is disabled with advice on what to change when it does not fit
• the loop checkbox writes LOOPSTART so nobody has to compute a sample offset
• the result can be previewed on the page, which for most people removes the question of what opens an .opus at all

@ffmpeg/ffmpeg and @ffmpeg/util come from npm and are imported lazily inside the click handler, so the static build never loads them. The 31 MB wasm core is fetched from a CDN rather than committed.

Checked

Ran the whole path against the production build, then fed the downloaded file to Emotecraft's own parser: mono, preSkip=312, LOOPSTART read back, 401 packets, round trip byte-identical.

Two things this turned up and fixed:

• plain VBR overshot a 32 kbps target by 30% (42 KB where the estimate said 32 KB), which matters when the ceiling is 1 MB. Encoding is now constrained VBR, and the estimate matches.
• <audio> never reported metadata in Chrome, so the duration comes from decodeAudioData instead — immediate and exact.

🤖 Generated with Claude Code

[cloudflare-workers-and-pages]

Deploying paldocs with    Cloudflare Pages

Latest commit:	6d8179e
Status:	✅  Deploy successful!
Preview URL:	https://f2ca4977.paldocs.pages.dev
Branch Preview URL:	https://feature-emote-sound-converte.paldocs.pages.dev

View logs