Help / Field writing

Voice capture not transcribing

Voice clips record on the device and transcribe in the cloud. Three things stop that from working.

Sign in to open this screen

1. Step 1

   Grant microphone permission

   The first time you tap Voice, the browser asks for microphone access. If you denied it, the button will appear but nothing records. On Chrome: Settings → Privacy → Microphone → allow inspecta.com. On iOS Safari: iPhone Settings → Safari → Microphone → Allow.

2. Step 2

   Clips queue without signal

   Voice clips record locally and upload to the transcription service when you have a connection. You will see a queued indicator in the writer. Stay on the page until the indicator clears, or reconnect to Wi-Fi when you leave the property. The clips do not disappear — they will upload and transcribe once the device is back online.

3. Step 3

   Transcript appears on the job, not the device

   The transcription runs in the cloud and posts back to the inspection on all devices. If you recorded on your phone but are reading on a laptop, reload the inspection page. It may take 30–60 seconds for the transcript to appear after the clip uploads.

4. Step 4

   Inspecta Copilot is required for Expand

   Holding Voice and recording a room produces a raw transcript. Expand — which splits the transcript onto the right systems and matches narratives — is part of Inspecta Copilot and is included on every plan. If Copilot is not configured on this workspace, the transcript appears as a note and you route it manually.

5. Step 5

   Voice transcription must be configured on the server

   Recording works on the device even when cloud transcription is off. If Voice saves a clip but returns “Voice transcription is not configured,” the workspace needs an Azure AI Foundry or OpenAI key. New Foundry resources use a https://YOUR-RESOURCE.services.ai.azure.com endpoint. Deploy gpt-4o-mini-transcribe (not Whisper, which is often unavailable) and set AI_TRANSCRIBE_MODEL to that deployment name.