okay sweet!! i'm not sure exactly what the technical limitations are (might follow your instructions and try this myself eventually? but i'm grabbing at straws for now lol), but i'd love to see something where light's voice gets more intensely emotional. there have been a couple instances in your work where his voice varies from like... his baseline, i guess? And that's been really cool to hear! Off the top of my head, the "i'm just exhausted, L" and "oh no, no, don't go, i don't mind" from the Light is tired. L is in love. clip were awesome.
if you're looking for a more specific setting, maybe nightmares about him or L getting taken out by kira during the yosuba arc? Or maybe nightmares about light killing L himself as kira, since that would play into what you did for the Light is tired. L is in love. clip (loved what you did there, btw–yotsuba!light feeling like he could actually be kira lives rent free in my brain). i could see light jerking awake in the middle of the night and frantically seeking L out for comfort–starting out totally panicked and desperate for L to confirm that they're both safe, and then settling into resigned dread that L tries to pull him out of. my own heavy preference for hurt/comfort aside (🤡), i'm really curious to see how the elevenlabs AI might handle pitching his voice higher and more desperate as he panics, and then down into a more dull, resigned tone when he's calmed down a little but still feels like shit–especially since there are so few samples of that in the show
thanks again for all your hard work on these! very excited to see your take on this if you choose to pick it up :D
This was actually a little tricky, so the clip itself sounds really rough and I decided to just leave it as is. The ai has trouble with setting the tone in shorter clips, more text means it follows the mood. When the clips are shorter there's more random varieties in the voice that are hard to wrangle, but I have faith that soon enough there will be more options to control it. In the meantime I had some luck with training based on moods of the voices so I'm going to dabble more with it and come back to this as a proper project. I really dig the idea and want to execute it better. I can get some squeaks and pitches in longer text but it's more unpredictable, so training the voice on samples that do that has also helped a little. I learnt a lot by playing with this so I really appreciate the request.
