Copy Diagnostics
Help › Copy Diagnostics (or Settings › About › Diagnostics › Copy) copies versions, settings, the last hour of errors and any crash in the last week to your clipboard, ready to paste somewhere. Read what you copied before you send it to anyone.
If Outmake quit unexpectedly
macOS keeps a report whenever an app stops unexpectedly. The next time Outmake opens, it says so once and offers Copy Report, a short summary of where in Outmake it stopped, without file paths or anything that identifies your Mac. Nothing is sent anywhere: you decide whether to paste it into an email.
Show Logs in Finder
Help › Show Logs in Finder opens the folder holding Outmake's logs, including install.log and the voice process's own voice.log. Settings › About › Logs › Open goes straight to transcript.log, what voice heard, said and did.
The model engine starting slowly
A model on this Mac isn't kept running between chats: the first message loads it, which can take a while for a large model, then it has to read your whole message before answering. Both are normal and get faster after the first reply. See Models on this Mac.
A model that won't load
If a local model fails to load, Outmake says why:
- "There isn't enough free memory for … right now." Quit apps you aren't using, or pick a smaller model.
- "macOS stopped the model engine, usually because memory ran short." Same fix: free up memory and try again.
- "… didn't load. Try again." A one-off failure. Copy Diagnostics has the details if it keeps happening.
A provider refusing a key
Pasting an API key (Settings › Models for images, Settings › Voice for OpenAI or Google speech) checks it right away and says why if it's refused, rather than only saying "saved" and leaving you to find out later.
For Claude, ChatGPT or Gemini, a chat's own error tells you what to do. Signed out of Claude or ChatGPT, the message has a Sign In button that starts the same sign-in Settings › Accounts does; for the others it opens Accounts. A usage limit says when it resets, and a connection problem asks you to check the internet and try again.
A picture OpenAI wouldn't make because the key is missing or wrong shows the key's field right under it. Paste a key there and the picture is made again.
Some models on Hugging Face are shared only with signed-in members. Their page in Models asks for a Hugging Face read token, which you make on huggingface.co under Settings › Access Tokens. Some makers also ask you to accept their terms on the model's own Hugging Face page before the download starts. The token is kept in your keychain and sent only to huggingface.co.