Jan AI
Jan is an open-source desktop AI assistant for running local models and connecting compatible cloud providers. Jan 0.8.4 adds native web search, an OpenAI-compatible Responses API gateway, and backend-managed settings with OS-keyring storage for credentials. Jan supports image uploads for visual understanding when the selected model can process images, but its official image-support documentation does not describe a built-in text-to-image generator.
People searching whether Jan can generate images can easily confuse the Jan.ai product with unrelated models that include “jan” in their owner or model name. The practical decision is whether you need a local assistant that can understand an uploaded image or a dedicated image-generation model that creates new images.
Jan documentation, changelog, and GitHub releases control product facts. Jan 0.8.4 documents native web search, Responses API translation, and a move from webview localStorage to a backend-managed settings store with OS-keyring credential storage. Jan's image-support changelog documents image upload and visual understanding with compatible multimodal models. The separate Replicate record `janb1956/jan` controls facts about that hosted image model and must not be treated as a Jan.ai feature source.
- Run supported language models locally through a desktop interface.
- Connect compatible remote model providers when local inference is not enough.
- Use native web search and fetch tools in supported Jan workflows.
- Upload screenshots or photos for visual understanding with a compatible model.
- Keep local-assistant and dedicated image-generation workflows clearly separated.
Jan can attach an image to a conversation and send it to a compatible vision model for analysis. That supports jobs such as describing a screenshot, extracting details, or reasoning about visual input. The result depends on the model and provider selected inside Jan.
Jan 0.8.4 adds native `web_search` and `web_fetch` tools, an OpenAI-compatible translating gateway with Responses API support, and per-model chat-template controls. It also moves provider settings and credentials from webview localStorage into a backend-managed store, with secrets stored in the operating-system keyring. Existing values migrate automatically, while the old localStorage data remains as a downgrade snapshot.
Uploading an image for a model to understand is different from creating a new image from a text prompt. Jan's official image-support announcement documents the former. If your job is text-to-image creation, use a documented image-generation model or service and verify its model ID, inputs, pricing, and license separately.
A Replicate model named `janb1956/jan` exposes image-generation and editing inputs, but the model owner and identifier do not make it part of Jan.ai. Product identity should come from the official Jan project and documentation, while hosted-model behavior should come from the specific provider record.
Jan AI FAQ
Common questions about Jan AI.
Can Jan AI generate images?+
Jan officially documents image upload and visual understanding with compatible multimodal models, not a built-in text-to-image generator. Use a dedicated image-generation model when you need to create new images, and do not treat the unrelated Replicate model `janb1956/jan` as a Jan.ai feature.
Why does GetLLMs also have a janb1956/jan model page?+
That page is a catalog record for a separately owned model hosted on Replicate. Its published image-generation inputs describe that model only; the similar name does not connect it to the Jan.ai desktop product.
Where does Jan store provider API keys?+
Starting with Jan 0.8.4, provider settings and credentials are stored through a backend-managed store, with secrets placed in the operating-system keyring. Existing values migrate automatically on first launch, while the previous localStorage data remains only as a downgrade snapshot.