Hello I'm having an issue with squads where regardless of how I configure my assistants (voice provider, assistant overrides or not), I have occurences of extremely slow speaking, as if the speed has been reduced by a ridiculous amount. Is this a known issue? It seems to not happen with transient assistants, but my issue with transient assistants is that I need to experiment with the system prompts, and all my tests with "PATCH" requests to change the system prompts end up deleting most of the squad.

Hi there! 👋 I understand you're experiencing two issues: extremely slow voice speed in Squads and problems with PATCH requests deleting squad configurations when trying to modify system prompts. I can help with both of these concerns. Regarding the slow voice speed in Squads, while I couldn't find information specifically about extremely slow speaking as you described, I did find several related voice issues in Squads: According to information I found, there are known voice-related issues in Squads, including: - Voice selection issues where users reported being unable to select a voice or the voice remaining stuck on "Sarah" [source1](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1333363724307660861), [source2](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1333379603028250665) - Voice distortion issues with 11labs voices that may be related to network issues, encoding/decoding errors, or problems with voice synthesis [source3](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1336377512233664523) - AI speech interruptions where the AI sometimes stops speaking mid-sentence [source11](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1334109401463783434) As for your issue with PATCH requests deleting squad configurations, this is indeed a known issue. According to the information I found: - A user reported that their tests with PATCH requests to change system prompts resulted in deleting most of the squad [source3](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1333363724307660861) - This happens because updating nested model properties via PATCH requests can overwrite properties such as the system prompt [source2](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1334944779850485790)

Well i'm using Vapi's Elliott voice, but no matter what voice I use, I still get extremely slow voices

Can you share the call ID?

@Shubham Bajaj Here's one : faeabdbc-3f7c-42cd-994f-28025c5d204d Listen to the audio at 1min16, this is what happens. I switched from using the API to doing my changes in the UI because my PATCH requests were overriding the entire squad. Maybe that's just the way I need to work? Keep overiding the whole squad and make changes directly in the JSON payload?

🔵 13:28:43:232 ElevenLabs (WebSocket #7) Connecting... URL: wss://api.elevenlabs.io/v1/text-to-speech/dN8hviqdNrAsEcL57yFj/stream-input?output_format=pcm_16000&optimize_streaming_latency=3&model_id=eleven_turbo_v2_5&enable_logging=false&enable_ssml_parsing=true 🔵 13:29:16:227 Voice Input Formatted: "I am your private reader. Before we start, I must ask a single question. Shall we begin?", Original: "I am your private reader. Before we start, I must ask a single question. Shall we begin?" 🔵 13:29:16:224 ElevenLabs (WebSocket #0) Connecting... URL: wss://api.elevenlabs.io/v1/text-to-speech/flq6f7yk4E4fJM5XTYuZ/stream-input?output_format=pcm_24000&optimize_streaming_latency=3&model_id=eleven_turbo_v2_5&enable_logging=false @Ouranos I just want you to confirm whether the second voice ID or the voice ID used for the assistant is intentional? Because it's all pure coming from 11Lab s side. I've looked into the logs. Nothing on our side. So wanted to confirm this with you. If you have more call IDs with the same behavior please share them as well because then it will help me push better your ticket. https://cdn.discordapp.com/attachments/1356259994647003167/1356748034116943993/Screenshot_2025-04-02_at_3.20.50_AM.png?ex=67edb192&is=67ec6012&hm=c4a0cb55c4cae603a6c68ad571ae287c8fd833c677463819a3e576e1a1874f76&

Hey @Shubham Bajaj so in the squad each assistant should have a different voice ID, if that's what you mean? The issue really occurs when I switch voices. This happens with other voice providers as well, Cartesia, Vapi etc

Hi, checking if this is resolved/solved for you?

@Shubham Bajaj no, still happening, I just updated my Squad via the API and I get this error. You can clearly see that the voice ID was properly picked up, but for some reason VAPI throws an error (specifically with Cartesia voices https://cdn.discordapp.com/attachments/1356259994647003167/1358736627911888977/image.png?ex=67f4ed98&is=67f39c18&hm=4df9f25494e966854a60c06a710c7f6b4534ebd492e86addf999aeba55b55f57&

@Shubham Bajaj I'm seeing this with a simple phone number to agent : 1 (808) 582-1177 My web interface has the expected voice ( https://dashboard.vapi.ai/assistants/1f357246-60da-49eb-a420-bfe686f835f2 ) but you'll see that the voice in that number, connected to this agent , is definitely slowed .

Kevin, Thanks for solving it all by yourself. Happy to help you further!

@Shubham Bajaj I started a new thread, it seems like some Cartesia voices are "theoretically" supported because they appear in the Cartesia community voices, but then they don't appear in the squads

Could you share your new thread link so I could understand better what exactly you are referring to?

@Shubham Bajaj https://discord.com/channels/1211482211119796234/1360533126039994389