Connect your own AI
Two ways in: paste an OpenAI key, or point Speakybara at a model you already run yourself on LM Studio, Ollama, llama.cpp or Unsloth/vLLM.
Explain with AI, Summarize, Translate, Improve writing, the optional AI clean-up and every action you write yourself need a model to talk to. Speakybara runs none of its own: you connect one under Models › AI model, on the right-hand side of the Models page. That is either OpenAI with your own key, or a model running on your own PC through LM Studio, Ollama, llama.cpp or Unsloth/vLLM.
You need Speakybara installed, and either an OpenAI account you can create a key in, or LM Studio, Ollama, llama.cpp or Unsloth/vLLM already running on your PC.
Connecting a model
Step 1
Open Models › AI model
The right-hand half of the Models page holds the AI fields: Provider, Base URL, Model, API key and a Test button. The left-hand half is about voices and has nothing to do with this.
Step 2
With OpenAI: pick it and paste your key
Choose OpenAI as the provider, paste your key into the API key field and press Test. The default model is gpt-4o-mini. The key is kept in config.json on your own PC and goes only to OpenAI, with each request.
Step 3
With a local model: pick your runner
Choose LM Studio, Ollama, llama.cpp or Unsloth/vLLM as the provider and the right address fills itself in: port 1234, 11434, 8080 and 8000 respectively. No key, and nothing goes over the internet.
Step 4
Not sure which model your server has loaded
Press Find models next to the model field. Speakybara asks your server for its list and you pick one; typing a name yourself is fine too.
Step 5
Run one action to be sure
Capture a sentence and try Explain with AI or Summarize. An answer means the connection works; nothing at all means the Test button on the Models page is the quickest way to see why.
Worth knowing
- What is sent is the recognised text, never the screenshot, as a chat request to the provider you chose. Speakybara has no server in that path.
- What OpenAI charges is between you and OpenAI. Speakybara takes no cut and adds nothing to it.
- A model you run yourself is separate software with its own appetite for memory and often for your graphics card, outside the roughly 55 MB Speakybara uses.
- One action of your own can use a different model from the rest, under that action's Advanced options.
If something goes wrong
It stays in config.json on your own computer and goes only to OpenAI, as the authorisation on each request. It is not sent anywhere else, and we never see it.
You set one provider at a time under Models › AI model, but a single action of your own can be pinned to a different model under its Advanced options.
Check that the runner — LM Studio, Ollama, llama.cpp or Unsloth/vLLM — is really running, and that the base URL matches its port. The app fills in the usual one, so a server on a different port has to be typed in by hand.