When using Gladia Solaria model, talking to assistant immediatly fails with error call.in-progress.error-vapifault-gladia-transcriber-failed Logs show that sample_rate of 24_000 is sent to Gladia API and the error message says sample_rate must be one of the following values: 8000, 16000, 32000, 44100, 48000" I managed to have this model work yesterday though and it now fails. Do you need more information to investigate? Is there a reliable workaround for this?

Thank you for the details. The Gladia API only accepts these sample rates: 8000, 16000, 32000, 44100, or **48000**—sending 24000 will cause an error. Vapi should automatically convert sample rates, but this may not be happening in your case. Could you confirm if you are explicitly setting the sample rate to 24000 anywhere in your configuration, or is this being set automatically by your integration? This will help determine if the issue is with your setup or a recent change in the platform. If needed, I can also guide you on configuring a fallback transcriber to avoid call failures. Sources: - [Vapi Gladia Transcriber Documentation](https://docs.vapi.ai/providers/transcriber/gladia) - [Vapi WebSocket Transport Documentation](https://docs.vapi.ai/calls/websocket-transport) - [Vapi Transcriber Fallback Plan](https://docs.vapi.ai/customization/transcriber-fallback-plan)

I did not set it manually

Hi Lucasg0, Thank you for reporting this issue with the Gladia Solaria model. Based on the error you're seeing, Vapi is sending an unsupported sample_rate (24,000) to Gladia's API, which only accepts 8000, 16000, 32000, 44100, or 48000. This appears to be a Vapi-side issue. To investigate and escalate this to our engineering team, I need: • Call ID (You can find this at [https://dashboard.vapi.ai/calls](https://dashboard.vapi.ai/calls)[)](https://dashboard.vapi.ai/calls)) • Timestamp of when the issue occurred Workaround options while we investigate: 1. Try a different Gladia model - Switch from

solaria-1

to

fast

or

accurate

to see if those models work with the current sample rate 2. Use a different transcriber provider - Deepgram, AssemblyAI, or Google STT are reliable alternatives Relevant documentation: • Transcriber fallback configuration: [https://docs.vapi.ai/customization/transcriber-fallback-plan](https://docs.vapi.ai/customization/transcriber-fallback-plan) • Gladia provider info: [https://docs.vapi.ai/providers/transcriber/gladia](https://docs.vapi.ai/providers/transcriber/gladia) Once I have the Call ID, I'll escalate this to engineering for a deeper review.