AI and models

Synthezia uses local Whisper and Ollama models in Local Mode. You can instead configure an OpenAI-compatible provider for supported AI operations.

Local

Local summarization

Send transcript text, summary instructions, and optional session context to the loopback Ollama runtime and persist the returned summary in SQLite.

RemoteNetwork required

Remote-provider summarization

Remote summarization transmits transcript and prompt context only after the user authorizes the normalized endpoint.

Model capabilities

Current model and provider capabilities

Capability	Status	Current behavior
Whisper model management	Available	List, download, cancel downloads for, select, and delete the Whisper models in the application catalog.
Ollama model management	Beta	List installed Ollama models and download, select, or delete models through the local Ollama runtime.
Machine-based model recommendations	Beta	Recommend Whisper and Ollama model defaults from detected memory, CPU count, and architecture.
OpenAI-compatible remote provider	Beta	Selecting and authorizing a remote provider can transmit transcript, summary, prompt, retrieved context, and conversation to that endpoint; embeddings remain local.

Model acquisition is a separate operation from local processing: downloads require a network connection even when you use Local Mode afterward. The privacy and security section explains the provider boundary in more detail.