David Holz told the world his company was profitable in August 2022. That was only a month after Midjourney launched into open beta. Now, sixteen months later, the company has released version 6 of its image generator. The money and the engineering have gone somewhere. The result is a clear jump in photorealism and in how closely the software follows what a user actually types.
Midjourney works like other AI image generators. You feed it a natural language description. It builds a picture from that text. OpenAI has DALL-E. Stability AI has Stable Diffusion. Midjourney has always competed on visual quality and on a distinct, often painterly style. Version 6 changes that calculus. The new release prioritizes photorealism. It also prioritizes prompt accuracy — meaning the image matches the description more precisely than before. For users who have watched the software misinterpret their words, that is the headline fact.
Holz co-founded Leap Motion, a company that tracked hand movements for virtual reality. That background in precise, physical interaction may explain Midjourney’s focus on detail. The company operates through Discord bot commands and through its official website. It does not have a public API like some competitors. It controls the user experience tightly. Version 6 suggests that tight control has paid off in engineering discipline.
The release date was December 21, 2023. That places it near the end of a year dominated by AI news. OpenAI had a boardroom crisis. Stability AI faced financial turbulence. Midjourney, by contrast, just kept shipping. The company announced profitability in August 2022, which is rare in a sector where most startups burn cash. Holz did not share revenue figures. He did not need to. The announcement itself signaled that Midjourney had found a paying customer base willing to subscribe for image generation.
That base has grown. Version 6 gives them a reason to stay. The photorealism improvement is not incremental. It is the kind of jump that changes what users expect from the tool. A designer can generate product mockups that look like photographs. An artist can render scenes that do not carry the usual telltale signs of AI generation — the warped hands, the nonsensical background details. Midjourney version 6 reduces those errors. It does not eliminate them, but it reduces them.
The broader context matters. Midjourney is one player in a boom. The AI image generation field did not exist three years ago. Now it has multiple major competitors, open-source alternatives, and a growing legal fight over training data. Midjourney has been sued by artists who claim their work was used without permission to train the model. That case is ongoing. Holz has not commented publicly on the specifics of the lawsuit. The company continues to develop.
Version 6 is the result of that development. It is a technical achievement. It is also a business move. Midjourney needs to stay ahead of DALL-E and Stable Diffusion. The release of version 6 suggests it has done so, at least for now. The improved prompt accuracy is the feature that will matter most to everyday users. A generator that does what you ask is a generator you trust. Trust keeps subscriptions active.
Holz started Midjourney with a small team. The company has grown but remains lean. It has not taken outside venture capital in large rounds like its competitors. Profitability gives it independence. Version 6 is the first major update since that profitability was announced. It validates the model. The company can fund its own research. It can set its own schedule. It released version 6 on a Thursday in late December. No fanfare. Just a better product.
