Original Discord Post by diepro1481 | 2024-08-21 08:17:28
Hi there, i’m trying to build a digital Character of a real person using Convai, Unreal Meta Humans and Elevenlabs for the Voice. I just wanted to check weather you can tweak the TTS system of Convai. When using the TTS intergrated in Elevenlabs the Voice sounds much more similar to the real person than in my digital Character. Seems like the Convai TTS forces the Voice to sound deeper and does not get the rythm right. Can i tweek the speed/pitch/energy of the Voice that is generated by Convais TTS? Or even integrate Elevenlabs-TTS to Unreal? Thanks for your help.
Hi <@1133141372501753857> Can you please tell us what parameters have you configured on the Elevenlabs UI. Parameters like: stability and similarity.
Also note that the version of Elevenlabs available on their playground uses their REST API because the entire text is available beforehand. Convai uses the websockets version of the Elevenlabs API which is meant for real-time usecases but has some discrepancy from the website version. Hope this helps!