Midjourney
AI image generation built for aesthetics, not photorealism
About Midjourney
Midjourney generates images from text prompts, and its reputation has always been aesthetic rather than technical: it produces work with a distinctive look instead of aiming purely at photorealism. The V8 series changed what the tool is capable of underneath that. Images now render at native 2K resolution in roughly five seconds, long and detailed prompts are followed far more closely than in earlier versions, and text placed in "quotes" inside a prompt is actually rendered rather than approximated into letter-shaped noise, which was the single most common complaint about the tool for years.
Two features define how people work with it now. Draft exploration lets you generate 24 variations of a prompt and pick which ones to render at full quality, which changes the economics of iteration: you are choosing from a spread rather than rerolling one image at a time. Personalization and moodboards learn your taste and bias output towards it, and V8.2 in particular is an aesthetics release built around that, with a larger pool of images feeding personalization profiles. Worth knowing before you commit: there is no free tier and no free trial, the cheapest plan is $10 a month, and pricing is no longer published on the website itself. The plan comparison lives in the help centre instead, which makes it harder than it should be to check what you are paying for.
You work through the web app or through Discord, and both drive the same account. Describe the image you want, including style, composition and mood, and put any text you want rendered inside quotes. The model returns a set of options, and with draft exploration you can widen that to 24 variations before deciding what deserves a full-quality render. From there you upscale, request variations, or refine towards what you had in mind. Personalization profiles and moodboards steer the aesthetic towards your preferences across sessions rather than requiring you to describe your taste in every prompt. Fast GPU hours are consumed per generation and vary by plan; Standard and above add unlimited generation in Relax mode, which trades speed for volume.
- •V8 Models - Native 2K output, roughly five second renders, and much closer adherence to long or detailed prompts
- •Text That Renders - Words placed in "quotes" inside a prompt appear as legible text rather than approximated shapes
- •Draft Exploration - Generate 24 variations of a prompt and choose which ones to render at full quality
- •Personalization and Moodboards - The model learns your taste and biases output towards it across sessions
- •Video Generation - SD on Basic, SD and HD from Standard upwards, with unlimited Relax video on Pro and Mega
- •Stealth Mode - Keeps your images and videos private, available only on the Pro and Mega plans
Artists, designers and concept artists who care more about a distinctive look than literal photorealism, and anyone producing visual work at volume where 24 drafts per prompt beats rerolling one at a time. The text rendering matters specifically to people making posters, thumbnails and mockups, which previously meant compositing type afterwards. Two things to check against your own situation. Stealth Mode is Pro and Mega only, so client work that cannot be public needs at least the $60 tier. And companies making over $1,000,000 a year in gross revenue are required by the terms to be on Pro or Mega regardless of usage, which is easy to miss when a $10 plan appears to cover you.
Pricing
$10 - $120/mo
Paid only, no free tier. Basic $10/month, Standard $30, Pro $60, Mega $120. Annual billing is 20% less ($96, $288, $576, $1,152 per year). Extra fast GPU time is $4/hour.
From the vendor pricing page, 2026-08-30
