Haven't seen this addon being mentioned here but I find it useful. It is called "OpenAI" (no, it is not made by OpenAI) and is available in the dev channel in the addons store. It requires, like the AI Content describer for example, an API key from either OpenAI, Mistral or OpenRouter (therefore, at least in the case of OpenAI, everything you send and recieve costs money).
It is a complete conversation interface to the AI models, and it is a great accessible alternative to the sometimes not that great web interfaces. It can also take a screenshot of either the screen or the current navigator object, as well as handling images from files or URLs (the latter a feature that is not that common) and send the image (or multiple images) with the request. A nifty side feature is that, when taking a screenshot, it saves it as an image file in the NVDA folder while the conversation is active, so if you want to e.g. get a "second opinion" with another AI you can copy that file somewhere else and have it described later.
There are some other nice features as well, like a customizable "system prompt" (a prompt that is sent to the AI where you can tell it how to respond, formatting, etc.), and transcribing speech from a microphone or audio files, the latter I haven't tested.
I haven't found many downsides, maybe that you can't get an image described straight away, you have to send it through the conversation dialog, however, that is just one keystroke away and you can even set a standard text specifically for images.
If you want to know more about it, here is the GitHub page:
https://github.com/aaclause/nvda-openai
By blindpk, 4 June, 2024
Forum
Windows
Comments
Been a minute...
It has been a minute since I had this add-on. Like the OP states, it is a conversation model, but once you have your API key up and running, it can be very useful. 😀