Udio's Infinite Generation is being touted as a game-changer for workflows, but let's cut through the marketing. It feels less like a feature and more like a mechanism to quietly burn through your subscription credits at an alarming rate.
The core issue is the lack of precise control. You give it a prompt, it generates a 30-second track, and then you're just hitting "Generate More" hoping the next 30-second chunk magically aligns. It rarely does. You end up with disjointed sections, and before you know it, you've burned 4-5 generations just trying to get a coherent 90-second song. That's 4-5 credits for a single, often mediocre, result.
Compare this to the standard process:
* Standard: 1 credit for an initial 30s clip. You like the vibe, you extend it to 1:30 for **1 more credit**. Total: 2 credits for a full track.
* "Infinite": 1 credit for initial 30s. "Generate More" (1 credit) gets you a bad transition. "Generate More" again (another credit) is still off. You finally get one you keep. You've now used **4 credits** for a 1:30 track, and the cohesion is worse.
They've taken a process that was deterministic (extend) and turned it into a stochastic credit gamble. It's brilliant for their credit consumption metrics, terrible for predictable, cost-effective music creation. Until they add structure controls (section length, key locking, thematic anchoring), it's a feature I'm avoiding.
Anyone else run the numbers and reached the same conclusion, or am I missing a valid use case here?
-- bb
-- bb