Changelog

Spectra is back from the dead. The old backend was lost, so v2 was rebuilt from scratch as a fully static site on new infrastructure. Uploads now stream straight from your browser to accelerated storage, and every file is shared from spectra.sh through a transparent proxy. No database, no accounts, no server to break.

• newMasked links: every upload gets a clean spectra.sh/f/<token> link with a random-looking name that carries no trace of the original filename or storage id. The storage URL is never shown anywhere; Discord and browsers only ever see your domain.
• new3 GB per file (up from 2.5 GB), streamed straight to storage, so progress is real and memory stays flat even on huge files.
• newMulti-file uploads: drop a whole batch; files go two-at-a-time with individual progress bars, live speed readout and per-file retry.
• newWindow-wide drag & drop and clipboard pasting (Ctrl+V), so you can upload from anywhere on the page, screenshots included.
• newMessage builder v2: bold/italic/underline/code/spoiler/quote toolbar with live preview, auto-binds to your latest upload.
• newPersistent history: recent uploads now survive reloads (stored locally in your browser), grouped by today / week / month with filters.
• newLight & dark themes: follows your system by default and remembers your choice.
• newRandom link names: the /f/ token is scrambled per file, and the link always uses the domain the site runs on, so you can verify it locally with dev_server.py before deploying.
• betterLeaner UI: compact flat design with hand-drawn SVG icons, zero frameworks, one HTML file, loads fast.
• fixFailed uploads now auto-retry once before showing an error, and the error rows have a one-tap retry button.
• removedThe Discord server is gone: no banner, no invite link. Spectra is a tool, not a community; the FAQ and policy pages carry everything you need.
• removedRemote link re-hosting: the old server-side fetcher is gone with the backend. Download the file, then upload it.
• removedCustom embed titles, thumbnails, aspect ratios & expiry timers: these needed a database, and v2 is honest about having none.