Skip to content

Kōrero v1.43.0 - see what's in use, a local model that fits, and long meetings read in full

Latest

Choose a tag to compare

@nkeating-cmd nkeating-cmd released this 02 Oct 10:10

Kia ora! 1.43 makes it easy to see what Kōrero is using, to run AI clean-up on your own computer with a model that fits it, and it fixes long meetings getting notes from only part of the transcript.

New in 1.43.0

  • Always see what's in use. The foot of the sidebar shows your microphone (the real device, even on "system default"), the speech model and the clean-up model with its prompt, on every page. Click any of them to change it. The clean-up line says where text goes: on this computer, or the service it's sent to. The microphone and speech model can't be changed mid-recording, and the menu says why.
  • A local AI model that fits your computer. AI clean-up & notes now opens with a card that reads your graphics card, memory and free disk, suggests one model that fits, and downloads it with progress and Cancel, test-runs it and switches to it. It installs or starts Ollama if needed. Other models are labelled Fits well, Fits, Slower or Won't fit. The suggestions haven't yet been tested on Kōrero's clean-up prompts, so treat them as a starting point.

Fixes

  • Long meetings no longer get notes from half the transcript. Ollama reads only 4,000 tokens by default on graphics cards under 24 GB, so notes and Ask for meetings longer than about 15–20 minutes were written from part of the meeting, without saying so. Kōrero now asks Ollama for a context sized to each request, refuses rather than cuts a transcript that can't fit, and keeps the model loaded at that size.
  • Meeting notes on a local model are much faster straight after a meeting. Kōrero's speech model and the AI model were sharing the graphics card, which made reading a long transcript about 20 times slower on a 12 GB laptop. Kōrero now frees the speech model's memory before a local model reads a long meeting (unless you've set the speech model to stay loaded). Your next dictation reloads it, which takes a few seconds.
  • Time limits for local models now allow for slower computers, so long notes finish instead of timing out.
  • The Today panel names your microphone instead of saying "Default".

Installing

Download Korero_1.43.0_x64-setup.exe below, or let Kōrero update itself. The installer isn't code-signed yet, so Windows SmartScreen may warn you: choose More info → Run anyway. Your settings, meetings, notes and recordings are kept.

Ngā mihi!