In this article, let's break it down, How to connect local AIs to automatic writing of articles in ClipAI. Let's consider the work of ClipAI with the program "LM Studio". If you wish, you can use almost any other program: the instructions will be approximately the same.

Customizing LM Studio

1. Download LM Studio from the official website. This program is completely free of charge.

2. Go through the installation and the first launch of LM Studio. During the first launch you will be prompted to download the AI model:

Most likely, you will be offered one of the simplest models. However, we recommend downloading it: it will be easier to customize the operation first, and then change the model to the one you need.

3. Select the downloaded model to use by clicking on the top "Select a model to load"and then selecting the desired AI:

4. Click the "Developer" button on the left side of the LM Studio window and then click "Start server":

Customizing the ClipAI

1. Go to the input settings of the ClipAI (right-click on the line with the script -> "Settings"). Then click the Options tab.

2. Enter the "Secret key" - any key (e.g. "ClipAI"). Also select the model "GPT (own model)":

3. Under "ChatGPT" below, enter:

• "Own model" - name of the model in LM Studio. You can see it either at the top of the program window or in the chat responses with AI. Remember, there should be no spaces in the model name. For example, in this case you need to enter "deepseek-r1-distill-qwen-7b":

• "Custom API Endpoint" is the address specified on the "Developer" -> "Reachable at" page of LM Studio, to which you need to add "/v1/chat/completions". Most likely, you will have "http://127.0.0.1:1234" specified on this page, which means you need to specify "http://127.0.0.1:1234/v1/chat/completions" in the ClipAI settings:

As a result, the settings in the "ChatGPT" section of the ClipAI program should look something like this:

Installing AI models in LM Studio

To download and install the AI model, all you need to do is click the "Discover" button on the left side of the LM Studio window:

After that, a window will open to search for and install various neural networks. Usually, the more it "weighs" and the newer it is, the better the quality of the articles. But don't forget that the size of the neural network should not exceed the size of your VRAM (video card memory) or, in the case of using it on Mac computers with m processors (e.g. macbook on m1), your free RAM (RAM).

Don't forget to change the value of the "Own Model" parameter in the "Parameters" tab of the ClipAI input settings after installing the new AI!

Most AI is optimized to work in English. We recommend using:

For Russian-language articles:

1. YandexGPT (e.g., yandexgpt-5-lite) - A good neural network for short articles.
2. Phi (e.g., phi-4) - is suitable for many languages, including Russian.
3. Magnum (e.g., magnum-v4-12b) - pretty well suited for a variety of tasks.
4. Mistral (e.g., mistral-nemo-instruct) - is very good at writing even long articles.
5. Qwen (e.g., qwen2.5-7b-instruct-1m) - often can't write a long article, but is quite good at writing short ones, usually without making mistakes in HTML markup.

For articles in Chinese:

DeepSeek (e.g., deepseek-r1-distill-qwen-32b). Produces a process of "thinking" before writing an article, which although it will not be added to the final text, but significantly affects the improvement of its quality. Writes even very long articles. We would like to add it to the list of recommended for Russian, but this AI often embeds Chinese characters or English words into the text.

Сообщение Как использовать ClipAI совместно с локальными ИИ (LLM)? появились сначала на ClipAI.

In this case, as an example, we will show the addition of the service GROQ. It is already added by default to ClipAI, so you do not need to repeat these steps specifically to use GROQ; if you need any other service, just repeat the steps below by analogy.

1. Find a OpenAI compatible URL in your neural network API documentation. For example, in GROQ it is located ­here:

2. Ibid. find the name of the model you need:

3. Create your API key and copy it. Specify it in the corresponding field of the "Parameters" tab of the input settings.

4. Now add all the copied parameters in the ClipAI input settings -> “Parameters” tab:

In this case, you need to specify the obtained "Model ID" in the "Custom Model" parameter and "base_url" in the "Custom API Endpoint" parameter.

And select the model “GPT (custom model)”:

Ready! Now you can use any neural network you need.

Сообщение Как добавить в ClipAI другую нейронную сеть? появились сначала на ClipAI.

Please note: almost all services are free to use.

	GPT (OpenAI)	Google Gemini	Together AI	Deepseek	OpenRouter, per account with a limit (60 requests per minute); there are completely free models that do not spend the balance (labeled "Free"), There are many completely free models (labeled "Free").	Mistral	Nvidia Cloud	GROQ
Quality (English)	It's great	It's great	Good	Good	Good	Good	Good	Good
Quality (Russian language)	It's great	It's great	Good	Average	Good	Average	Good	Good
Free to use	Appears after selecting ­here "Enabled for all projects."	Unlimited free use, capped (more than 60 requests per minute and 14'950 per day total for multiple models, including Gemma) per account; or free $300 per Google Cloud account, but virtually no limits.	Got it полностью бесплатные models that do not spend the balance (labeled "Free"), but are available only after entering the card data (without debiting).	—	Присутствует множество полностью бесплатных models (labeled "Free"). Up to 50 free requests per day per account immediately after registration, or up to 1000 requests per day - if the account has ever been funded by $10.	Provided free 1 million tokens (~2500 articles) per month per account.	Provided free of charge up to 40 requests per minute to the account. The validity period of the API key is about 6 months.	Provided free of charge from 250 to 14400 requests per day on each of the model offered by the service.
Regions of operation (IP)	Except RF/RB	Except RF/RB. IP addresses are supported more than 210 countries. EU IPs may sometimes not be supported.

• The table shows the characteristics (quality, cost) of the highest quality model if the service provides several models.
• The "Quality" rows in the table reflect the subjective opinion of the ClipAI team, based in part on user opinion.
• “RF” - Russia, “RB” - Belarus, “EU” - European Union.
• The "Work regions" line reflects from which IP addresses of which countries the request for article generation can be sent. Additional restrictions may also be imposed on the part of the service used, for example, in the usage rules.

This article contains information relevant to 07.10.2025. We update it as necessary (if any changes occur in the world of AI). Most likely, everything described above is still relevant today. At the same time ClipAI updated constantly.

Сообщение Какой ИИ выбрать для написания статей? появились сначала на ClipAI.

Let’s look at the settings that are located in the “Advanced Settings” section. We recommend changing them only if you know exactly what you are doing.

“Timeout waiting for a response to a request from AI (sec)”. The maximum time to wait for a response to each request. A larger value will reduce the number of errors when there is no response from the server or an empty response is received. At the same time, sometimes this can cause text to take longer to generate. Default is 360 sec.

"Repetition rate". A lower value may cause various problems in the article (random character set, transitions to other languages, and so on). However, sometimes this helps reduce the Spam parameter and expand the AI's vocabulary. Default is 5 points.

"Variety of text". A lower value can reduce the uniqueness of the text and make it “dry”, as well as increase the number of logical repetitions. However, AI will also become more likely to use those facts in which it is confident. Default - 5 points; and for some AIs - 3 points.

"Custom API Endpoint". Link to connect to an alternative API compatible with the GPT API when working with AI. Needed, for example, if you are using local AI, like GPT4All. Or use a general (for example, corporate) Reverse Proxy. The default value is https://api.openai.com/v1/chat/completions.

Сообщение Расширенные параметры появились сначала на ClipAI.