Choose your models
Outcome
Section titled “Outcome”You have an appropriate local transcription model and local summary model selected for the workflows you intend to use.
The three model roles
Section titled “The three model roles”- A Whisper model turns audio into a local transcript.
- An Ollama model creates local summaries and answers chat questions in Local Mode.
- The local search model improves semantic retrieval for Global Chat. It can be requested separately from the summary model.
- Open Models from the sidebar or press Cmd+M.
- Under Transcription, download an available Whisper model or select an installed one as the default.
- Under Summarization & Chat, select an installed Ollama model or expand Download more models and download one.
- Open Settings to choose the default summary mode and control whether recommended local models are prepared in the background.
- Run a short non-sensitive recording to decide whether the chosen model is practical for your machine and content.
Expected result
Section titled “Expected result”New imports and recordings use the selected Whisper default. Local summaries and chat use the selected Ollama model when Local Mode is active.
Important limits
Section titled “Important limits”Synthezia uses detected memory, CPU count, and architecture to recommend defaults. Those are heuristics, not accuracy, speed, compatibility, or capacity guarantees. Catalog download sizes are estimates and can change.