3.10 AI Parameters and Synthesis
Consonant duration, pitch, multi-dimensional AI parameters, and the vocals you hear are all synthesized by AI. In order to achieve this uncompromising AI technology and maintain high-performance vocal expression, we utilize cloud synthesising to handle the high-intensity computing demand. After you hit play, you will need to wait briefly before hearing the completed synthesis, during which time you can cancel playback at any time. The synthesising task will continue in the cloud, or you can cancel it directly to edit the clip. Of course, we are continually optimizing the cloud synthesising and reducing the wait time.
During synthesising, the MIDI and pieces being synthesized or queued for synthesising will blink, and the blinking will stop once synthesis is completed. Synthesized pieces are displayed in dark color, while pieces that are not rendered remain in light color.
In AI synthesis, each layer of AI reasoning requires sufficient contextual information while also considering rendering flexibility. Therefore, we introduce synthesis pieces, which means that we divide each MIDI segment into several small vocal pieces based on the note gap size. These pieces are synthesized one after another. You can see the gray-colored shapes on the time ruler of the piano window, which are the smallest vocal units used for synthesis.
If you notice:
- Stuttering in the middle of a sentence, you can check if the pieces are split here and adjust the note gap to change the segmentation result.
- Unstable vocals between two sentences, you can also check if the pieces are split at the point of change and adjust the note gap to merge these two pieces.