stability tip - short vs long messages

[potent sage]

Maybe post them here? @lone valley

[lone valley]

good idea, I'll do that, thanks

well here's the 9 example sound files

3 are considered a short message, which resulted in a stability setting of 0.85 with the following message: Hello
3 are considered a longer message (mid), which resulted in a stability setting of 0.71~: Good morning. How are you?
3 are considered a long message, which resulted in a stability setting of 0.35: Keep in mind that the audio will be played through the virtual output device, and other applications that are listening to the virtual microphone will receive the audio

I'm not sure if these are the kind of examples you're looking for, if you want me to adapt something else then let me know (subject to my character usage that is) 👍

the main reason I came up with this kind of interpolation idea for the stability setting is because previously I'd be saying something really short like hello with a low stability setting and every so often it would be a rather unrealistic kind of hello based on my cloned voice. Obviously the examples aren't my voice, just a generated one 😂

[potent sage]

Great tip, thanks for sharing. Will definitely keep this in mind as I'm about to start coding a project using the API. Personally have barely used short inputs yet through the website, but my application may need to.

[lone valley]

Nice, yeah my use case is probably a bit unique

I don't often have access to a microphone, so I figured what about cloning my voice and appearing to still talk on voice calls in such a situation. Thanks to ElevenLabs I'm now managing to do that 🙂

[potent sage]

Yep, one reason I prefer TTS is that the workflow is so much more efficient including not needing to have a mic and appropriate place to record without background noise.