I've been using Suno for a few weeks now to generate some audio assets for a side project. While I appreciate the core model's capabilities, I can't help but focus on the frontend performance. The interface feels noticeably sluggish, which impacts the iterative workflow.
From a backend perspective, I'm curious about the underlying architecture. A few specific pain points:
* **Input Lag:** There's a perceptible delay between typing in the prompt and the UI responding. This suggests either excessive JavaScript bundle sizes or unoptimized event handling on the frontend.
* **Asset Loading:** After generation, loading the audio preview and the final file often feels slower than the actual model inference time. This points to potential issues with their CDN strategy or how they're serving static assets.
* **State Management:** Navigating between previous generations and the main interface sometimes triggers a full re-render of components, instead of a smooth, cached transition.
I wonder if they're serving a monolithic frontend bundle. A more modern, component-driven approach with code-splitting could significantly improve the perceived performance. For heavy, interactive apps like this, consider:
```javascript
// Example: Lazy load non-critical UI components
const AudioPlayer = React.lazy(() => import('./AudioPlayer'));
```
Has anyone else done a deep dive into the network tab while using it? I'd be interested to see waterfall charts or Lighthouse scores. The latency in the UI makes the creative process feel less fluid than it should be.
-- latency
sub-100ms or bust