Voice in v2.5 broken

[letswastetimee]

Anyone else having issues with voice after upgrading to v2.5? Mine sometimes speaks a Sim-like language or OVERemphasizes words. It's fairly creepy and kind of absurd.

[cybrhawk101]

Mine wasn't working as expected. It never connected to ElevenLabs and I finally asked PAI to find a local solution to run on my MAC. It found and build a working voice using Kokoro TTS model.

[MarkAC007]

Yes, same issue. I asked PAI to troubleshoot and fix. 👌

[emory]

my DA is based on Peter Capaldi's performance as The 12th Doctor in BBC's Doctor Who* and uses a voice I made with ElevenLabs.

My upgrade to 2.5 was a big jump, I was running with a pre-1.0 environment longer than I care to admit, and I was seeing a lot of YOUR_VOICE_ID_HERE placeholders.

I went looking and wound up realizing my core contexts didn't get picked up, so once I realized that I told the DA they were missing and what they contained and we should be using <pasted id> for a voice and it got sorted out.

I still don't have a speech interface I like, but at least when I am at that house I can hear announcements 😩

*: My PAI actually identifies as the TARDIS but brings 12th Doctor Energy.

[danielmiessler]

This was a voice server configuration bug in v2.5 that's been fixed in v3.0. The voice server now validates the voice_id field on startup and the phase announcement curl commands use explicit IDs. Upgrade to v3.0 and it should be resolved.