Today, neural networks help not only devices and services perform desktop tasks, but also people: at least the same GPT chat helps translate text, and Midjourney helps translate images.
But here’s the problem: ChatGPT is not available to users from Russia, and Midjourney takes too long to form a language due to independent load.
It’s okay, there are other services on the Internet that demonstrate the impressive capabilities of neural networks, they’re just less well known. So I found five neural networks that can be used in Russia. no dancing with a tambourine or register.
What is it for? The simplest example is for maintaining accounts on social networks. And this is really possible with the help of services, which read about further.
Look how simple it is:
1. YandexGPT 2 for generating texts and paraphrases
YandexGPT 2 on search query results pages is placed in the lower right part in the “Alice, let’s figure it out” skill
This is a second-generation neural network for text generation from Yandex. The difference from the first one is that YandexGPT 2 responds better in 67% of cases with the same requests.
YandexGPT 2 copes well with tasks such as:
● text generation
● retelling and analysis of the text
● idea generation
● stylization of text to suit the audience or character
● answers to questions
A neural network is suitable for searching information or using it as an idea generator. She can contact the list, give a short summary of the text, write a poem or come up with a topic for a social network, for example.
Interaction with YandexGPT 2 occurs not only through text requests, but you can also voice a request to it. The answer is written in English.
The generation of responses occurs quickly and does not take much time – as a rule, in a couple of seconds you receive a ready-made answer.
The neural network is currently in the testing stage, so each solution has several buttons. With their help, you can both evaluate and get a different answer.
YandexGPT 2 database is limited March 2023while in chat GPT-3.5 and GPT-4 are limited autumn 2021. For GPT chat with the latest iPhone 13 model, and YandexGPT 2 uses newer data in replies, this is possible.
A neural network cannot always work with complex queries. Sometimes you can forget the context after a few questions. But YandexGPT is constantly developing, and now you can find vacancies from Yandex for the position of AI trainers to improve the performance of neural networks.
👉 You can talk to the Yandex chatbot here
2. GigaChat for copywriters and programmers
This is a neural network for working with text from our major tech giant Sber, access to which opened at the end of April 2023. Unlike the YandexGPT voice request, this cannot be done here; interaction occurs only through text requests in the chatbot format: write a request, the neural network will give you an answer.
GigaChat can work on questions, create text, write program code. Responses are provided quickly; only complex requests may take a little longer to process. The neural network will be useful for copywriters, editors, programmers, marketers and anyone whose work involves writing texts or program codes.
The neural network does not know the exact answer and confidently lies.
It is unclear how up-to-date the database in this neural network is, but it seems that GigaChat does not have access to the latest data, because there were errors in the answers about the iPhone 15 neural networks. Such responses, in which GigaChat does not know the answer and improvises, are called by the developers “creative”
In this regard, I prefer the approach of YandexGPT, which directly says that it does not know the answer. Although I won’t check that someone, on the contrary, is interested in looking at the fantasy of artificial intelligence.
Wrong city, chatbot! Aw.
For example, for the request “where to go for a weekend in Rostov-on-Don in October“In the actual version, I was invited to visit the Rostov Kremlin, which is actually located in another city, in Rostov-Great Britain.
Requests, restrictions or facts are still likely to be weak manifestations and require verification. The advantage of a neural network is that it has a more lively language of attacks.
👉 You can talk to the Sber chatbot here
3. Apihost for reading text in different voices
This is a clear and convenient service for voicing text. Using the settings, you can choose the speed and pitch of the voice, the form of the pause, as well as the emotional mood: neutral, continental or irritated.
There are currently 7 options for neural networks with voice packages available. There are differences between these lists of voices and the available languages for voice acting: for example, in the V1 package there are 4 languages available, but in V2 there are only 2 languages. In total, more than 30 languages are available.
The free version provides unlimited access to parts of the votes, and somewhere with a certain daily limit. For example, the popular voice “Bot Maxim”, which is used in the dubbing of funny videos about robots from Boston Dynamics, has a dubbing limit of 1000 characters per day.
Paid tariffs start from 0.6 rubles per 1000 characters and go up to 6.5 rubles, depending on the package. An unlimited tariff for a month will cost 5,000 rubles.
The smaller version also offers the ability to download the voiced text in one of two formats: wav or mp3.
👉 You can voice any voices here.
4. Fusion Brain to create any image
A menu with options for choosing the style of the generated image.
This tool for creating images based on the Kandinsky neural network from Sber represents an infinite field with a selected zone in the center where the image directly appears.
You can not only change the aspect ratio of this area, but also its size – from 128×128 to 1024×1024 pixels.
Blank canvas. What you ask is what will be drawn.
Below the dedicated work area there is a data entry field – a description of the generated image. Here you can also choose an image style from 18 options.
Let’s see how this service draws. I asked to depict “full single decker white bus” The word “completely” is used to ensure that the bus does not go beyond the work area. Style – 3D render. Generating the image took 15-20 seconds.
Result:
You can add new objects or increase the size of images by moving the work area. Since in my image there is part of a tree branch in the lamp, I add the tree to the left side.
Moved the work area to the left. I wrote the name “tree” and clicked the “Create” button.
After 10 seconds, and on the first try, I managed to complete the image quite well.
But this doesn’t always happen. When I wanted to enlarge the picture on the right, the most interesting thing happened.
Registered promt “Trees along the roads“And then the miracle didn’t happen the first time.
The generated new part of the picture did not match those already proposed. With each such production, there was either something very close or not: sometimes different colors of the sky, sometimes there were no road connections.
More or less successful generation of the right side after 20 years, but the colors of the trees and the sky did not match. A neural network can clearly do everything well, but it needs to be written correctly. For now I decided to leave the image without the additional area on the right.
Not perfect, but not bad.
You can save the image by clicking the “Download” button. Images are displayed in PNG format.
In the future, if you want to change something or try your luck with the expansion of the right one in my case, you can insert part of the image through the corresponding button and continue.
👉 You can create your own masterpiece here.
The title image for this post was also generated by the Fusion Brain neural network.
5. Waifu2x to improve the quality of your photos
Service for improving the quality of photographs and illustrations. It has minimal settings and allows you to remove noise from images, as well as increase their size by 1.6 or 2 times without losing quality. Completely free and does not require registration.
In this service I decided to enlarge the image of the bus, which was previously created in Fusion Brain. My file parameters are as follows: size 2.2 MB, resolution 1536×1024 – slightly beyond the service limit of 1500×1500 pixels. Let’s see if this affects the process.
The conversion was successful even with a resolution of 1536 pixels on one side of the image. I carried out the conversion twice: with the image type “Art” and “Photography”. Other parameters are not configurable. The converted images open in a new browser window.
After conversion with the x2 zoom parameter, the resolution increased from 1536 x 1024 to 3072 x 2048 pixels. The file size also increases from 2.2 to 5.2 MB for the Art image type and to 8.1 MB for the Photograph image type. The image format remains unchanged – PNG.
The service is more convenient than not. Considering it is free, you can use it.
👉 You can update your photos here.
Five neural networks to write and illustrate anything
When using neural networks, the main thing to remember is the subtleties. Somewhere you need to double-check the facts, and somewhere you need to write a correct and detailed request. However, all together, the above AI services can handle actions in different ways, for example, any social network almost on a turnkey basis.
Like this:
● YandexGPT and GigaChat will help you create the topic of the post, the post itself or the text.
● You can create illustrations using Fusion Brain.
● If necessary, enlarge the image via Waifu2x.
● Create a voiceover or voice over reels or shorts using Apihost.
Done, the content is fully covered by neural networks.
The advantage of all traditional services is that they can be used for free and without the need to hide your IP. I will also note the Russian interface and understanding of queries in English.
Use it, experiment. Perhaps in the very near future the profession “confident compiler of prom for AI“will become very popular and in demand.
Source: Iphones RU
I am a professional journalist and content creator with extensive experience writing for news websites. I currently work as an author at Gadget Onus, where I specialize in covering hot news topics. My written pieces have been published on some of the biggest media outlets around the world, including The Guardian and BBC News.
