ACE-Step 1.5: Colab setup and custom UI tutorial

This tutorial covers two Google Colab notebooks for ACE-Step 1.5. One launches the official Gradio interface; the other uses the speaker's custom UI. Both run on a hosted T4 GPU.

Setup starts with selecting the T4 runtime and choosing Run all. The default notebook checks the environment, installs dependencies and downloads model weights before providing a Gradio URL. The speaker initializes the service, then explains generation with prompts, supplied lyrics and presets. The Think toggle lets the language model refine the input before audio generation. The speaker recommends disabling it when the prompt and lyrics are already detailed.

The custom interface has create, cover, remix and repaint tabs. Generation controls include language, instrumental mode, duration up to 480 seconds and a seed setting. The speaker reports generating two one-minute tracks in under two minutes with the demonstrated T4 configuration.

For repainting, the speaker uploads a generated song, enters its lyrics and selects a timeline region to change. The reported results are mixed: short word or phrase replacements can work after retries, while longer sentence replacements may need further attempts.