Tunova

Changelog

What shipped, and why — newest first. We build in the open so you can see the thing is alive and moving.

Breaking changes to the request/response shape get at least 30 days notice — see the versioning policy.

reliability

Cold-start protection keeps the first request after idle warm

Continuous background checks now keep the service warm and catch any degraded state before a real request reaches it — so the first call after a quiet period is as reliable as any other, not a cold start. (And as always, any render that does fail is auto-refunded.)

content

Playable proof + honest comparison guides

The landing now plays the actual mp3 a sample call returned, and two new guides compare the real ways to reach Suno (self-host, generalist hubs, and Suno-focused services) with the trade-offs stated plainly.

api

Rate limit surfaced proactively

When limiter state is available, successful generation/lyrics accepts and 429 responses carry RateLimit-* headers (Reset is a 0–60 second delay); during fail-open degradation those quota headers are omitted. GET /api/balance reports your configured limit and remaining token balance.

sdk

SDKs v0.1.0 — Python + Node, zero dependencies

Published open-source (MIT) clients for Python and Node/TypeScript with a built-in HMAC webhook verifier and an MCP manifest, at github.com/erliona/tunova-sdk. Copy-paste or install — your call; the raw REST API stays curl-simple.

api

v5.5 selectable per request

The newest Suno model is selectable per request via the model parameter, and billed only on a successful render — a failed generation refunds itself on the same statement.