Midjourney

Midjourney is an image generator based on a text description (prompt). Works via Discord and web interface. In terms of the quality of the “artistic” result, it consistently outperforms its competitors: the pictures look like concept art from a design studio, and not like computer graphics.

In my practice, Midjourney has taken the place of the primary tool for creating article covers, advertising banners and moodboards. In 5–10 prompts I get options that previously required 2–3 hours of designer work and a brief. It is especially valuable that the model understands stylistic references: “in the style of Leica street photography” or “editorial, natural light” - and produces exactly that.

A disadvantage that is important for the Russian market: payment only works with foreign cards or through intermediaries. The second disadvantage is that prompts must be written in English; Midjourney doesn’t understand Russian well. To generate blog covers, I use the following combination: concept text → GPT-4 translates into a detailed English prompt → Midjourney generates → easy modification in Figma.