Let me play a TV show, The House. It's not The House, it's just House. Ah? Okay. House. Also. The last. Video. Wait. It's the second one. This one? So you're recording this. It's still going. So I want to show this to you, so basically you have a kind of autoscroll. So I can differentiate between speakers. Yeah, I'm not good at it. Oh, it's really good. Yeah, so what do you use for transcription? It's just DeepGram. DeepGram has diarization, which I believe is actually part of the model. They were able to also diarize Whisper, so it's some sort of rapper-speaker-ish model. It works okay if people are speaking in... But what I've observed is because it's... It takes out previous context. If it misses once, it starts missing more until it sort of stands back up. Both of you are speaking like... Yeah, yeah. What's it doing now? In the interest of saving time and avoiding a lot of boring chit-chat later, I'm Dr. Gregory House. You can call me Greg. I'm one of three doctors staffing this clinic this morning. Short, sweet, gratifying. This ray of sunshine is Dr. Lisa Cuddy. Dr. Cuddy runs this whole hospital, so unfortunately she's much too busy to deal with you. I am a board... I kind of want a scrolling bar. I suppose people want to add a scrolling bar when it's over one page so people know how long it goes. Interesting. The scrolling bar is absent. I think it was added yesterday, but it's gone right now. How long can this go? Be forever? You'll see if they're still going. If the light's blue, it's still going. That one never failed me. We can stop the record so I can show Andrew. The transcription is always running. Is it on the phone? These two are separate. Yeah, so the way this works is the transcriptions get... These get... The audio gets streamed in backend. Then this is a diarized preview, but transcribing the whole file works better because the model has more context. So afterwards, the summaries are done by re-transcribing the whole file. Let me see if it shows.