Short answer
Midjourney V7 is a completely new image model that Midjourney opened for alpha testing on April 4, 2025, its first new model in nearly a year. It follows text prompts more closely, draws hands, bodies and objects more coherently, turns personalization on by default, and adds a Draft Mode that makes rough images ten times faster at half the cost.
Follow-up question How do you start using Midjourney V7?
Yesterday, April 4, 2025, Midjourney began letting its community test V7, the seventh major version of its image generator. The rollout started around midnight Eastern time, and it is Midjourney's first new model in nearly a year. The company calls it an alpha: an early test version, released so users can find its strengths and weaknesses before it is finished.
Midjourney turns a written description, called a prompt, into images. Its founder, David Holz, announced V7 on the company's Discord server, describing it as smarter with text prompts, better with image prompts, sharper in quality and texture, and much more coherent in the details of bodies, hands and objects.
What changed in V7
Midjourney stresses that V7 is an entirely new model, not an update of V6, and warns that it may need a different style of prompting. Three changes matter most.
Personalization is on by default. Before you can use V7, you build a personalization profile by rating images, about 200 of them according to TechCrunch, which Midjourney says takes roughly five minutes. The model then leans toward the kinds of images you said you like. You can switch it off at any time.
Draft Mode makes quick, rougher images for exploring ideas, at ten times the speed and half the cost of a normal job.
New modes and prices. V7 launches in two speeds, Turbo and Relax. A Turbo job costs twice as much as a normal V6 job; the standard speed is promised soon. Some tools, including upscaling (making an image larger and sharper), editing and retexturing, still use the V6 models for now. Midjourney says to expect new features every week or two for about two months, the largest being a new way to keep a character or object consistent across images.
How personalization works
Two people can type "a cozy reading corner" and picture very different rooms: one imagines a sunny Scandinavian nook, the other a dark library full of leather chairs. A prompt never says everything, so the model has to fill the gaps with its own guesses.
Personalization gives it better guesses. By rating images, you show the model what you find beautiful: colors, lighting, styles, levels of detail. Midjourney then uses that profile whenever your prompt leaves room for interpretation. The company argues this helps it understand both what you want and what you find attractive.
The trade-off is that results become partly about you. If you want a neutral starting point, or you are making images for someone with different taste, turn personalization off.
How Draft Mode works
Draft Mode is like a designer's thumbnail sketches. Instead of finishing every idea, you produce many quick versions, pick the promising ones, and only then spend effort on the final piece.
On the website, turning on Draft Mode changes the prompt bar into a conversational mode. You can type an instruction such as "swap the cat for an owl" or "make it night", and Midjourney rewrites the prompt and starts a new job for you. A microphone button adds a voice mode, so you can speak your changes aloud. When a draft looks right, choosing "enhance" or "vary" re-renders it at full quality. Midjourney says drafts are lower in quality but behave consistently with the final renders, so what you choose is what you get, only sharper.
Here is what a draft session might look like:
| Step | What you do | What happens |
|---|---|---|
| 1 | Type a short prompt with Draft Mode on | Rough versions appear quickly |
| 2 | Say or type "make it a rainy evening" | The prompt is rewritten and new drafts start |
| 3 | Say "add a red umbrella in the foreground" | Another round of drafts |
| 4 | Click "enhance" on your favorite | It is rendered again at full quality |
Why it matters
V7 arrives a week after OpenAI built image generation directly into ChatGPT, a feature that spread fast online. Midjourney's answer is not a chat assistant but a dedicated artist's tool that tries to learn your taste and make experimenting cheap. Draft Mode in particular changes the rhythm of the work: when an idea costs half as much and appears ten times faster, people try more of them.
Midjourney is also an unusual company. Holz, who co-founded the hardware maker Leap Motion, started it in 2022, and TechCrunch reports it has never raised outside money. Like other image-model makers, it faces lawsuits from artists who say their work was used for training without consent, a debate that V7 does not settle.
Try it: Pick one scene, such as "a lighthouse on a cliff". Generate it in V7 once with personalization on and once with it off, using the same prompt. Compare the two sets and write down three differences, such as color, mood or framing. Those differences show you what your profile has learned.
How do you start using Midjourney V7?
On the Midjourney website, look for the Version setting and pick V7 from its list. If you have not built a personalization profile yet, you will be asked to rate images first; set aside about five minutes. Members of Midjourney's Discord server can switch to V7 there too.
Start with Draft Mode to explore. You can also add --draft to the end of a prompt to run a draft job directly:
a lighthouse on a cliff at dusk, waves crashing, warm light in the window, watercolor style --draftBecause V7 is new, older prompts may behave differently. Keep prompts clear and concrete, describe the subject first and the style second, and adjust one thing at a time so you can see what each change does. When you find a draft you like, enhance it, and remember that Turbo jobs cost more while Relax is slower but cheaper.
Sources
- V7 Alpha — Midjourney, April 4, 2025
- Midjourney releases V7, its first new AI image model in nearly a year — TechCrunch, April 3, 2025