AI-generated Answer for rafailmpalis_40203
Hello! I understand you're experiencing an issue with text-to-speech using websockets where the voice isn't asking the user's question as expected. This could be related to a known issue with certain voices when using websockets, particularly with the Turbo v2.5 model.
To address this problem, here are a few suggestions:
-
Voice selection: Ensure that the voice you're using is fine-tuned for the Turbo v2.5 model. You can verify this in the My Voices section or via the API.
-
Model choice: If you're generating content solely in English, consider using the Turbo v2 model instead. It offers similar performance and doesn't exhibit the same issues as v2.5.
-
Chunk length: Review your minimum chunk length setting. The recommended default is between 120 to 290. Using a setting outside this range can result in lower audio quality.
-
Voice quality: For better stability and accuracy, especially with longer voiceovers, try using one of the high-quality professional voice clones. These are marked with an HQ tag in the Voice Library.
-
Text length: Keep your prompts to 1500 characters or less, particularly when using synthetic voices (marked with a flask icon) or instant voice clones.
If you're still experiencing issues after trying these suggestions, it would be helpful to know which specific voice and model you're using, and whether this problem occurs consistently or only with certain voices and models.
pseudo shadow
pale meteor