How to generate AI vocals from existing vocal audio?
In ACE Studio, use stem splitter to separate vocal tracks, convert vocal to MIDI & lyrics, edit melodies and lyrics if needed, then generate.
Unlike RVC-type AI vocal generators, ACE Studio is not merely a simple AI voice changer. Instead, it converts vocal audio into MIDI and lyrics, which are then performed by an AI singer. This allows for a high degree of editability in both the lyrics and melody of the vocal.
Steps:
- Convert vocal into MIDI & lyrics
- Separate vocal tracks from accompaniment (if need)
- Edit lyrics
- Generate AI vocal
Convert vocal into MIDI & lyrics
If you have a vocal track without any accompaniment:
Simply drag it into the audio track in ACE Studio. Then, click the 'Vocal to MIDI & Lyrics' button.
After the conversion is complete, you will get a MIDI project with lyrics.
If your audio file contains both vocals and accompaniment：
Drag the audio file into the audio track of ACE Studio, then click the "Split into stems" button to separate the vocals from the accompaniment.
Next, select the vocal track and proceed as described in the "If you have a vocal track without any accompaniment..." section.
Edit lyrics
The automatically recognized lyrics might look a bit strange because the AI model actually recognizes phonemes and then infers the corresponding words from them. No worries, let's just click the play button to listen to the AI singer's performance.
If there are any mispronunciations, you can correct them by modifying the corresponding lyrics. Double-click on a specific note to modify the lyrics corresponding to that note.
Alternatively, as you can see that the lyrics of connected notes are grouped together to form a phrase, you can directly edit the entire phrase here.
Please note that each note can only carry one phoneme.
- If you input a multi-syllable word on a single note, the syllables will be automatically distributed across several consecutive notes. For example, "paradise" has three syllables, and it will be distributed across three notes, represented as paradise#1, paradise#2, and paradise#3 to indicate the respective syllables.
- If you want multiple notes to sing a single-syllable word, you can input the lyrics on the first note and then input a sustain symbol "-" on the following notes. For example, if you want to use 3 notes to sing the word "hours," you can do it like this:
Generate AI vocal
Click the play button to hear the AI singer perform.
Drag another AI singer from the voice list to hear different versions of the performance.
Click here to export audio.
