Skip to main content

How do I use ClipAI with local AI models (LLMs)?

In most cases, you don't need a local AI at all!

In ClipAI, it is almost always best to use high-quality AI models that are completely free (such as Google Gemini). To use them, you don't need to follow the steps in this guide.

To use local AI you need:

1. A PC with a graphics card with at least 8 GB of video memory, or an Apple computer with an M-series chip (such as M1).
2. The skills of an experienced PC user.

In this article, we explain how to connect local AI models to ClipAI for automatic article writing. As an example, we use ClipAI with the “LM Studio” app. You can use almost any other program if you like: the steps will be roughly the same.

Setting up LM Studio

1. Download LM Studio from the official website. The program is completely free.

2. Install and launch LM Studio for the first time. During the first launch, you will be asked to download an AI model:

First-launch window with model download in LM Studio

Most likely, you will be offered one of the simplest models. However, we recommend downloading it: it is easier to get everything working first and then switch to the model you need.

3. Select the downloaded model: at the top, click “Select a model to load”, then choose the AI you need:

Button for choosing the model to load in LM Studio

4. On the left side of the LM Studio window, click “Developer”, then “Start server”:

“Developer” section button in LM Studio
“Developer” button
“Start server” button in LM Studio that starts the OpenAI-compatible REST API
“Start server” switch

Setting up ClipAI

1. Open the ClipAI input settings (right-click the line with the script -> “Settings”). Then open the “Options” tab.

2. Enter any “Secret key” (for example, “ClipAI”). Also select the “GPT (own model)” model:

Setting up ClipAI with LM Studio
The “Options” tab of the ClipAI input settings

3. Below, in the “ChatGPT” section, enter:

  • “Own model”: the name of the model in LM Studio. You can find it at the top of the program window or in the AI's chat replies. Note that the model name must not contain spaces. For example, here you need to enter “deepseek-r1-distill-qwen-7b”:
Model name at the top of the LM Studio window
Model name at the top of the LM Studio window
Model name in the LM Studio chat
Model name in the LM Studio chat
  • “Custom API Endpoint”: the address shown in LM Studio under “Developer” -> “Reachable at”, with “/v1/chat/completions” added to the end. Most likely, it shows “http://127.0.0.1:1234”, which means you need to enter “http://127.0.0.1:1234/v1/chat/completions” in the ClipAI settings:
LM Studio API address
LM Studio API address

As a result, the “ChatGPT” section of the ClipAI input settings should look roughly like this:

The “ChatGPT” section of the ClipAI input settings
The “ChatGPT” section of the ClipAI input settings
If ClipAI and LM Studio are installed on different PCs
If ClipAI and LM Studio are installed on different PCs in the same local network, or if you run ClipAI in Parallels Desktop and LM Studio in macOS, you need to:

  1. In LM Studio, click the “Settings” button to the right of the “Start server” switch in the “Developer” section and enable “Serve on Local Network”.
  2. Check the address shown just to the right (“Reachable at”). It should have changed. Enter this address in ClipAI instead of “127.0.0.1”. Be sure to add “/v1/chat/completions” at the end.
  3. In some cases (only if nothing works after the first two steps) you also need to allow the port in Windows Firewall on the computer where LM Studio is installed. The default port is “1234”; open it for both TCP and UDP. You don't need to do this on the computer with ClipAI. You also don't need to open the port on your router unless the computer with LM Studio and the computer with ClipAI are connected to different networks (Internet connections).

Installing AI models in LM Studio

To download and install an AI model, just click the “Discover” button on the left side of the LM Studio window:

“Discover” button in LM Studio
“Discover” button for downloading models in LM Studio

This opens a window where you can search for and install various neural networks. Usually, the larger the model and the newer it is, the better the articles. But remember that the model must not be larger than your VRAM (graphics card memory) or, on Mac computers with M-series chips (such as a MacBook with M1), your free RAM.

After installing a new AI, don't forget to change the “Own model” value on the “Options” tab of the ClipAI input settings!

Which models to choose

Most models write best in English. For articles in Russian in November 2026, we tested Gemma 4 31B: it writes well and can handle long articles, but it requires a graphics card with about 24 GB of memory or a Mac with 32 GB or more. On less powerful machines, try the smaller models from the Gemma, Qwen, and Mistral families.

Please note that Gemma is available even without a powerful computer: for free, using the same Google AI Studio key as Gemini. Running it locally makes sense if you need to work without an internet connection or without daily limits.

Still have questions? Contact @a1exmurz or use the online chat (button in the bottom-right corner). We will answer any questions at any time. We can also set up ClipAI on your computer (via remote access) tailored to your needs. It's free!

ru_RURussian