Hi Vercel community!

I'm building KI Song Erstellen (KlangKI), a German-first AI music app for people who want to turn an idea or their own lyrics into a song without learning music production.

Try it: https://kisongerstellen.com/

What you can do:

• Generate vocal songs from a text description or your own lyrics, or create instrumental tracks.
• Generate structured German or English lyrics, then use them for a song.
• Adjust style, mood, instruments and vocal preferences.
• Separate uploaded audio into vocals and accompaniment, or four stems.
• Listen to results in your library, with downloads and public song sharing available to paid users.

On the implementation side, the app uses Next.js App Router, TypeScript, Tailwind CSS, next-intl, Better Auth, Drizzle/PostgreSQL and Cloudflare R2. German is the default locale, with an English version as well. The music generator is embedded directly in the landing page.

The workflow is asynchronous: generation runs as a task, the UI shows its status, and completed tracks become available for playback. Failed tasks have an explicit state, with credit refunds where applicable. Keeping that waiting experience understandable is a big part of the product.

I'd welcome feedback on the first-time experience, especially how clearly the page explains the difference between starting with a description and starting with existing lyrics. If you've built a Next.js app around long-running AI tasks, I'd also love to hear what has worked well for your progress and recovery UI.

Thanks for taking a look!