Why is only my voice in the transcript?

Call audio isn't being captured. It needs macOS 14.2 or later and the Screen & System Audio Recording permission, followed by a full restart of Oyma.

Why doesn't anything happen when I ask for a summary?

Usually there's no AI model available. With default settings, summaries need Ollama running on your Mac with a model pulled. Otherwise add your own API key and route summaries to the cloud.

Does the diagnostics file include my notes?

No. It contains app and macOS versions, your settings with secrets removed, and recent logs with paths, email addresses and API keys scrubbed. It never includes note content.

Troubleshooting

Most problems come down to a missing permission, a model that isn’t downloaded, or no AI provider being set up. Start with the built-in check in Settings › Meetings › Diagnostics: click Run check and it tests your microphone, call audio and speech model in about five seconds.

Call audio is silent, or only my voice is transcribed

The other participants come through the sound your Mac plays, which needs:

macOS 14.2 or later. Check in Apple menu › About This Mac. On older versions only your microphone is recorded.
Screen & System Audio Recording permission. In Settings › Meetings › Permissions, the Meeting audio row should read Allowed. If it reads Needs setup, click Open System Settings and turn Oyma on.
A full restart. After granting the permission, quit with ⌘Q and reopen. Until then the row reads Restart required.
The setting itself. Transcribe other participants in Settings › Meetings must be on. It is by default.

Then click Run check while something is playing. Connected, no sound for meeting audio can just mean nothing was playing during the check. Full steps are in Recording permissions on macOS.

If your microphone shows In use by another app, close the other app and try again. If it shows Connected, no sound, check the input level in System Settings › Sound and that the microphone isn’t muted.

Words appear twice in the transcript

Your microphone is hearing the call through your speakers while call audio is also captured. Oyma doesn’t cancel echo, so wear headphones.

Transcription is slow or lags behind speech

First use of a model. Small and Large v3 Turbo take minutes to prepare the first time they load. Live text lags until that finishes, and Run check shows Still preparing. Base is ready in seconds.
Running without the GPU. If Run check shows Ready (CPU — slower), transcription is running on the processor and text lands a few seconds behind speech. This usually means the graphics version of the model couldn’t be downloaded. Make sure the model shows as downloaded under Model on this device, and if Local-only mode is on, download it from there once.
Model size. Larger models are more accurate but slower. Base is the fastest. See Transcription models and languages.

“Recording saved. Set up the speech model…”

Transcribing a file and re-transcribing a meeting use the Base model. In Settings › Meetings, select Base under Speech model, click Download, then run >Re-transcribe current meeting from the command palette. You can switch the live model back afterwards.

The summary doesn’t appear, or shows an error

By default, summaries and Ask run on your Mac through Ollama, which Oyma doesn’t install. Check, in order:

Is an AI model set up? On the AI tab, Meeting summary & chat shows where they run. If it reads On device, you need Ollama running with a model pulled. If it reads Cloud, you need an API key. Set up either one with Summaries and Ask with Ollama or your own key.
Is the summary turned off? In Settings › Privacy, under Per-feature routing, Meeting summary must not be Off. Summarise after transcription in Settings › Meetings controls the automatic summary.
Does the note say ai: false? Notes with that frontmatter are never summarised.
Is the transcript empty? There’s nothing to summarise if no speech was recognized.

Errors you might see:

“No AI provider is configured…”: summaries are routed to the cloud with no provider set. Add a key in AI settings, or set Meeting summary to On device.
“On-device AI isn’t available yet…”: Oyma can’t reach Ollama. See the next section.

To retry, open Summary from the waveform button and click Try again or Generate summary, or run >Re-summarise current meeting (AI) from the command palette.

Ollama isn’t found

Is Ollama running? Look for its icon in the menu bar. If it’s missing, open Ollama from Applications.
Is the model pulled? In Terminal, ollama list shows your models. The one selected for Ollama (local) in AI settings must be listed. If it isn’t, run ollama pull llama3.2, or the model you chose.
Test it. In AI settings, click Test connection in the Ollama row. Connected means Oyma can reach it.
Is it slow rather than missing? On Apple silicon, an Intel build of Ollama runs under Rosetta without the GPU. file "$(which ollama)" should mention arm64. Reinstalling from ollama.com fixes it.

Meetings aren’t detected

Detection reads window titles, which needs the Screen & System Audio Recording permission. If you see Enable meeting detection, grant the permission and restart Oyma. The supported apps are listed in Record a meeting. You can always start capture by hand with ⌘⇧R.

Export diagnostics for support

Open Settings › Privacy and scroll to Support. Or press ⌘O and run >Export diagnostics….
Click Export diagnostics… and choose where to save the file.
Send the Markdown file to hello@getoyma.com with a short description of what happened.

The file contains app and macOS versions, your settings with secrets left out, and the most recent app log, with file paths, email addresses and API keys scrubbed. It never includes note content. It’s a plain text file, so you can read it before sending.