Hey, don't need help, (part of teh fun is figuring the shit out 🙂 ) - once I adjusted the Transcriber away from using same provider (openAI in this case) all started working as expected again. Funnily, the issue sarted after I incorporated fallBack transcriber, and at that same time, I decided to test if openAI transcriber is more reliable. was using 11labs to that point, but it failed a few times, due to service outage, so I learnt about fallBack setup.

{
  "transcriber": {
    "model": "whisper",
    "language": "en",
    "provider": "talkscriber",
    "fallbackPlan": {
      "transcribers": [
        {
          "provider": "11labs",
          "language": "en",
          "model": "scribe_v1"
        }
      ]
    }
  }
}

is my current AI transcriber setup. I am building out transient AI's each personalised per call, to the caller, identified by number. I am testing out different transcribers as 11labs had a few outages, and also seems to struggle a bit with some of the Australian accents on calls. Waiting to hear from client testing how whisper is doing. I am currently busy building out tests/test environment that incorporates local unitTests, and voice tests via (again transient) test Suite runs, from which I have 'known input/client data setup' -> voice test -> unit tests on initial data to expected known outcomes. (like registering a minor on the system attached to calling parent) It is at this point (as per by bug raised on support channel) that I started noticing transcriber badness