Add your own voice to songs

[ivaxsirc]

Custom Vocal Integration (Voice Input + Voice Cloning)

I would like to propose adding a “Custom Vocal Integration” feature to ACE-Step 1.5, with two complementary modes:

1. Voice Performance Mode

Allow users to upload a recorded vocal performance (e.g., singing or humming), and use it as the main vocal track of the final song.

The system would:

Detect tempo, key, and structure from the input audio
Generate full instrumentation around the user’s voice (drums, bass, harmonies, etc.)
Preserve the original vocal recording without replacing it

This would be a natural extension of the existing Extract/Lego/Complete workflow.

2. Voice Cloning Mode

Allow users to provide a short reference of their voice so the system can generate entirely new songs sung in a similar voice.

The system would:

Learn vocal characteristics from a short audio sample
Generate new melodies and lyrics performed in the user’s voice
Optionally allow style control (emotion, intensity, genre adaptation)
Why this is useful
Makes music generation more personal and expressive
Bridges the gap between AI composition and human performance
Expands ACE-Step from a “music generator” into a true collaborative tool between user and model

I believe even implementing the first mode would already be a major step forward, while the second would unlock a much deeper level of creative control.