Co-founder of Kettik Group, Bakdaulet Bekzhanov, on how AI is changing content production

Bakdaulet Bekzhanov, Co-founder of Kettik Group, Instagram

About Me

I create content using artificial intelligence. I have a media background—I am the co-founder of Kettik Group, formerly known as "Let's Go Out." For the past three years, I have been studying generative AI and applying it to content creation. I also share knowledge with colleagues and other companies, explaining how AI works and how to use it effectively in content production.

About the Product and Technology

Technologies have already become part of everyday life—we are surrounded by ChatGPT, Gemini, and other AI tools. If you perceive AI as another intelligence that you can communicate with like a human, it becomes easier and clearer how to interact with it.

Generative AI is one of the key innovations of recent years. It has sparked a wave of new startups and solutions actively developing in this field.

About Closed Source

Among Closed Source technologies, there are graphic models. We will mainly focus on graphic and video models. Text models are not of particular interest in the current context. As an example of a Closed Source graphic model, we can mention Midjourney, which is familiar to many. There are also closed-source video models—one of the most popular today is Sora from OpenAI. All these solutions are Closed Source. Language models, where text is input and text is output, are a separate type that we are not considering at this time. Audio models will also not be discussed in the current context.

About Open Source

Open Source technologies have begun to develop rapidly, and if we look at them from a historical perspective, in some cases they have demonstrated even more significant progress than Closed Source solutions—such as, for example, DALL·E 3. A bright example is the company Stable Diffusion, which made a real breakthrough and achieved a high level of controllability over generation results.

Stable Diffusion, as an Open Source project, proved the effectiveness of the open approach. Later, a team split off from it and launched a project called Flux, developed by Black Forest Labs. It is known to the general public as Flux—a model that provided extremely realistic image generation. This was a turning point when it became difficult to distinguish AI-generated images from real photos.

When the Flux 1.1 Pro version was released, it became obvious that we were entering a new era of visual technologies. Very recently, the company released a new model—Flux Kontext. Its main difference is support for so-called "omni-editing," where image editing is performed using text instructions. The built-in language model understands commands and allows changes to be made within the image, ensuring a high level of interactivity and flexibility.

Flux Kontext has become a serious competitor to multimodal Closed Source solutions, such as GPT-4 from OpenAI, which also gained the ability to generate and edit images. However, OpenAI has not yet managed to maintain object identity between editing sessions—stability and consistency of the image were lost. Flux Kontext, on the other hand, ensures a high degree of consistency, which is especially important when creating characters, brands, or advertising visuals, where visual discrepancies are unacceptable.

For example, imagine we have a character—a seagull in VR glasses. With Flux Kontext, we can place it in any context. We control the object's position, its emotions, and the environment—this solves one of the key tasks of generation: controlling the image and its consistency across different scenes.

If you want to try Flux yourself, keep in mind that the model is resource-intensive. It is desirable to have at least an RTX 40 series graphics card, otherwise the process will be too slow. Even with powerful equipment, many prefer to use cloud solutions for the sake of time.

For those who want to delve deeper into Open Source technologies, you can use the ComfyUI interface. This is a node-based system that allows you to visually configure and combine generative models, create pipelines for generating, editing, and post-processing images. ComfyUI allows you to use Flux and other models in a single workspace.

There is also the Freepik Generate Images service, where you can run graphic models directly in the browser. The system uses internal currency, equivalent to the consumption of cloud resources, and allows you to experiment with different generation options. Omni-editing is also available here, which gives the opportunity to specify changes in Russian, with English supported as the primary processing logic.

About Application

If we talk about professions, any specialists involved in content creation will undoubtedly find use for these technologies. The capabilities of generative models allow creating impressive visual effects, personalized images, selecting clothing styles, and using avatars for voiceover and character visualization. For example, I personally created a version where I appear as a news anchor and talk about my educational course. This is just one of the options for use.

At the same time, the scope of application of such solutions goes far beyond the online space. Tools are already actively used in television, video production, and the film industry. In my case, the technology helps in creating a feature-length film based on AI—the project is too early to discuss in detail, but it could be one of the first films fully generated using artificial intelligence.

Flux Kontext is just one of the tools actively used within a broader process. In content production, it is important to know how to combine and use many solutions: from generating images to their processing and final editing. It does not end with Flux Context—it is part of the technology stack.

There are already hundreds, if not thousands, of image generation models on the market. Among the best at the moment, we can highlight Imagine 4, Midjourney, Flux, as well as the generative capabilities of ChatGPT based on Sora. These tools are actively used by SMM specialists, marketers, targetologists, creators, videographers, directors, advertising agencies, and other specialists working with visual content.