This new product from NVIDIA allows you to create your own ChatGPT and run it from a PC.

It’s now easier to run your own ChatGPT locally on your PC thanks to NVIDIA. The company launched Chat with RTX, an app that makes it easy to create a chatbot by taking advantage of your computer’s hardware. Of course, if it is equipped with a proprietary graphics processor.

As NVIDIA explained, the chat with RTX is a demonstration. This means that the software is not yet 100% complete and may contain errors. However, it is a very interesting alternative that can help speed up tasks such as searching for specific data in documents stored on a PC or creating summaries of videos uploaded to YouTube.

The application has a minimalistic interface and its operation is far from complicated. Once installed, users can choose between two Large Language Models (LLMs) for running Chat with RTX: Mistralfrom the French startup of the same name, or FLAME 2from Meta.

Subsequently, they need to select the folder containing the documents they want to scan using the NVIDIA software and that’s it. From now on, they will be able to ask questions to the chatbot, just like when using apps like ChatGPT, Gemini or Copilot. Artificial intelligence will be responsible for analyzing the files provided to it and extracting answers from there.

So, for example, if Chat with RTX analyzes a folder of files related to your next vacation, you can ask it to tell you the day and time of your flight. NVIDIA Chatbot will not only give you a text-based response, but it will also link the documents it extracted data from so you can quickly open them.

Continuing with the same example, if you already have an activity schedule for your trip and it’s in PDF format, you can ask it to extract that information and summarize it for you in seconds. But Chat with RTX also has the ability to work with items that aren’t stored locally. like the video on YouTube.

NVIDIA lets you create your own ChatGPT using Chat with RTX

If, instead of using local data, you want to extract information from videos hosted on YouTube, Chat with RTX allows you to do this seamlessly. You only need copy the link and paste it into the NVIDIA application. The AI ​​will be responsible for processing the content so that you can ask it questions. Additionally, as with documents, you can create a resume and offer it in text format.

In the promotional advertisement for this software, NVIDIA mentions that the app works with “selected videos.” Although this is not discussed in detail, it is logical to assume that the results differ depending on the content. Simply put: Don’t be surprised if your app crashes frequently when using this feature.

As we said at the beginning, Chat with RTX requires an NVIDIA GPU in your computer. The chatbot is compatible with all models in the series. PTX 30 and PTX 40 which have at least 8 GB of video memory. Other requirements to install it are Windows 11, at least 16 GB of RAM, and video card drivers version 535.11 or higher.

Although it is likely that you will get real benefit from the application you need a fairly powerful computer. As Tom Warren noted, Edge, the application takes up about 40 GB of memory and uses a Python instance that consumes about 3 GB of RAM when running. Finally, it’s worth clarifying that if you use an application with local data, said information will not be uploaded to the cloud or leave your computer.