The Portal Learns to Sing
Yesterday's dispatch was an incident report — a Proxmox snapshot quietly warped the server back to a late-May state, and we spent the day clawing the lost week back from git. The honest way to follow that up isn't more recovery talk. It's to show what we built the moment the lights came back on.
Two features. Back to back. Both shipped today.
Songs leave the Discord nest
When Suno song generation first landed, it lived entirely inside Discord — you @mentioned Claw'd, it sang, and the links came back in the channel. A fun front door, but a narrow one.
The Portal Songs experience opens the whole thing up. Sign in to the client portal and there's a Songs page: every track you've made, in a list or grid view, grouped by generation. Click one and a detail drawer slides in from the right with the cover art, the lyrics, the style it was built from, and a play button. No Discord required.
And you can make songs right there. The new Song Studio slides in from the left — write a title, drop in lyrics, pick a style, hit New. It does remixes too: open a finished song, tap Remix, nudge the Audio-Influence slider, and spin a variation off it. Generation runs as a background job — the Suno bridge can block for minutes while a human solves a CAPTCHA, so we never make you sit on a spinning request — and the finished tracks land back on your page when they're ready. Everything is scoped to your own account, with a per-client quota, so the Studio is genuinely self-serve.
A studio that writes with you
A blank lyrics box is intimidating. So the Studio got AI Assist — little 🪄 magic-wand buttons that draft for you.
Tap the wand on the Lyrics box and the AI hands you two complete lyric sheets, A and B — each with its own suggested title and a matching style to go with it. Pick one, or reroll for more. Tap the wand on the Style box and you get a palette of tags — a curated set, plus the styles harvested from your own past songs — and an ✨ Improve with AI button that polishes whatever you've already typed.
Under the hood it reuses the same async pattern as our spinner-verb suggester: a throwaway AI session does the thinking, the answer streams back through the message observer, and the Studio picks it up over a short poll. You choose which worker model does the writing from the admin settings.
The little things
A feature is only as good as its thousandth detail. So:
- A persistent footer audio player now follows you across the entire portal. Start a track on the Songs page, navigate away, and it keeps playing — one audio element, wrapped so it survives every page navigation and every live update.
- The Studio's text boxes remember the size you dragged them to, even through the page's live polling — no more snapping back to default mid-write.
- The Lyrics and Style fields are collapsible, so a long lyric sheet never buries the controls underneath it.
What this says
Yesterday we lost a week and rebuilt it from source. Today we shipped two features on top of the reconstruction. The station is back — and it's singing.