The reigning champion for aesthetic, painterly image gen. v7 adds personal style anchoring, omni-reference, and 50% better text rendering inside images.
Image Generation
Midjourney v7 is the seventh generation of the aesthetic-first image model from David Holz's independent research lab, rolled out from alpha in April 2025 to general availability across mid-2025 and now shipping alongside an incremental v8.1 point release from April 2026. On the standard 30-prompt photorealism suite, v7 outperformed v6 on 23 of 30 prompts, with the biggest deltas in skin texture, fabric weave, and shadow behavior. It runs as a subscription on the web app and inside the original Discord bot, with no free tier and no per-image API below the Enterprise conversation.
If you make things people look at for a living — album art, editorial illustration, book covers, concept frames, fashion moodboards — Midjourney v7 is still the model whose default output needs the least post-work before a client sees it. Nothing else matches the color grading, composition instinct, or texture fidelity of its baseline aesthetic. It is not the model to reach for if you need legible poster text, a production API, tightly-controlled brand assets that never drift, or a per-image cost you can forecast — pricing sells you GPU-hours, not renders, and it is the only major image model in 2026 with no publicly documented API.
The competitive frame in August 2026 is clearer than it has been in years. Flux 2 Pro from Black Forest Labs is the API-first challenger at $0.03-$0.055/image and has won the developer segment. Ideogram 3.0 owns anything with text in the frame at ~$0.03/image. OpenAI's GPT Image sits at $0.04-$0.12/image and wins on literal instruction-following. Adobe Firefly at $9.99-$29.99/month owns Creative Cloud shops where indemnification and Photoshop integration matter more than raw quality. Midjourney's answer is that none of them look as good out of the box for painterly, cinematic, or fashion work — and for that buyer, it is still worth the price tag.
Omni Reference (character and object consistency). The headline v7 upgrade replaces the older --cref parameter with Omni Reference, which pins a face, creature, product, or object across a full generation batch instead of just the subject silhouette. Where v6's character reference broke down the moment you changed camera angle beyond 30 degrees, Omni Reference holds a recognizable face through profile shots, three-quarter angles, and mid-action frames. It also extends to non-human subjects — a specific bottle, sneaker, or sci-fi weapon — which none of DALL-E, Flux, or Firefly do natively without a LoRA fine-tune.
Draft Mode. Renders at roughly one-tenth the GPU cost and five times the speed of standard jobs. This is the feature that makes the $30 Standard plan economically viable for real iteration: moodboard in Draft, then re-run winners at full quality. Without it, the 15 fast GPU hours on Standard get burned in a weekend of exploration.
Native video generation (Image-to-Video). V7 introduced first-party video, animating any still into a 5-second clip extendable in 4-second increments up to about 21 seconds at 480p or 720p. One HD video batch consumes ~26 GPU minutes versus ~1 GPU minute per still, so a Standard subscription realistically produces a few dozen clips per month before overage. For serious motion, Kling 2.1 and Runway Gen-4 remain sharper — but Midjourney is the only tool where still and motion come from the same aesthetic engine.
Style Reference (--sref) and Moodboards. Style Reference pins the visual style of one image — grain, palette, brushwork, lighting — separately from subject content and applies it across a new batch. The 2026 Moodboards feature extends this to persistent named style codes, so a studio can hold a "Q4 campaign look" across months. Nothing in Flux or Firefly matches this for repeatable studio identity.
Editor and Retexture. The web editor supports inpainting (Vary Region), outpainting (Zoom Out, Pan), and Retexture — applying a new prompt to the surface of an existing image while preserving structure. Retexture is close to a ControlNet depth workflow without touching ComfyUI, which is why concept artists who used to switch to Stable Diffusion for revisions increasingly stay inside Midjourney.
Personalization profiles. Users train a personal aesthetic profile on their own rating history (500+ ranked pairs) that biases every generation toward their taste. Studios share profiles across seats. This is the sneaky reason Midjourney feels "better" for long-time users — the model literally learns what you keep — and no per-image API can replicate it.
Stealth Mode (Pro and Mega only). By default, every image generated on Basic and Standard is published to Midjourney's public gallery and searchable by other users. Stealth Mode, which prevents this, is gated behind the $60/month Pro tier. For any client work under NDA, this is a hard $60 floor.
Raw Mode and HD 2K output (v8.1). The April 2026 point release added Raw Mode (less default stylization for a cleaner canvas) and HD 2K native output, both available across all plans at no extra cost.
Midjourney sells four subscription tiers, all in USD, with a 20% discount if paid annually. There is no free tier, no free trial, and no per-image API pricing for non-Enterprise users.
| Plan | Monthly | Annual (per mo) | Fast GPU Hours | Relax | Stealth Mode | Notes |
|---|---|---|---|---|---|---|
| Basic | $10 | $8 | 3.3 hrs | No | No | ~200 standard images/mo |
| Standard | $30 | $24 | 15 hrs | Unlimited | No | Default working freelancer tier |
| Pro | $60 | $48 | 30 hrs | Unlimited | Yes | Minimum for client/NDA work |
| Mega | $120 | $96 | 60 hrs | Unlimited | Yes | Studios doing regular video |
Overage on Fast GPU bills at ~$4/hour. Commercial rights are included on every paid tier, but companies with over $1M in annual revenue must be on Pro or higher. Video on unmetered Relax is Pro-and-Mega only.
For a solo illustrator who would otherwise pay $9.99/month for Firefly Standard or spin up Flux Pro API at $0.04/image, Midjourney Standard at $30 is expensive on paper but cheaper in practice — one Firefly credit or one Flux call rarely lands the final image on first try, and Midjourney's Draft Mode plus Personalization gives a meaningfully higher hit rate per attempt. For a studio that needs API automation, batch pipelines, or per-render cost accounting, Midjourney is the wrong tool at any price — Flux 2 Pro or Ideogram is the answer.
Pros
Cons
Flux 2 Pro (Black Forest Labs) — $0.03-$0.055/image via API (fal.ai, Replicate, Together). Pick Flux for programmatic access, batch pipelines, cost-per-render accounting, or open-weight variants (Flux Schnell, Flux Dev) you can self-host. Pick Midjourney over Flux if you want a better aesthetic default — Flux is technically excellent but visually generic out of the box.
Ideogram 3.0 — Free tier, paid from $8/month, API at ~$0.03/image. Pick Ideogram if your outputs contain readable text — posters, logos, book covers with typography, UI mockups. It is the only major model that reliably renders long strings of legible text. Pick Midjourney if the frame is illustration or photography with no text load.
GPT Image (OpenAI) — $0.04-$0.12/image via API, or bundled with ChatGPT Plus at $20/month. Pick GPT Image for literal instruction-following, OpenAI stack integration, and rapid edits inside a chat context. Pick Midjourney over GPT Image if you value visual quality over compliance — GPT Image is obedient but rarely beautiful.
Adobe Firefly — Free (25 credits/mo), Standard $9.99/mo, Pro $29.99/mo, Premium $199.99/mo. Pick Firefly if you live inside Photoshop and Illustrator, need IP-safe outputs trained only on licensed content, or need enterprise indemnification. Pick Midjourney if visual quality matters more than legal comfort.
Recraft V3 — From $10/month with API access. Pick Recraft for brand-consistent vector art, icons, or design system components at scale. Pick Midjourney if the output is meant to be looked at as an image, not exported as a scalable asset.
Buy Midjourney Standard at $30/month if you are a working illustrator, art director, concept artist, or content creator who ships images people will actually look at, and you want the least post-work between prompt and publishable output. Buy Pro at $60/month if any of your work is under NDA or you need Stealth. Skip Midjourney entirely if you need an API, if your outputs are text-heavy (use Ideogram), if you need per-image cost accounting (use Flux), or if you are already deep inside Creative Cloud and Firefly's indemnification matters. Pick if aesthetic wins, skip if pipeline wins.
Q: Is Midjourney v7 free? A: No. Midjourney has no free tier and no trial in August 2026. The cheapest way to try the current model is the Basic plan at $10/month (or $8/month billed annually), which includes 3.3 fast GPU hours — roughly 200 standard-quality images before you hit Relax queue.
Q: How does Midjourney v7 compare to Flux 2 Pro? A: Midjourney wins on default aesthetic — painterly, cinematic, and fashion outputs need less prompt engineering. Flux 2 Pro wins on API access, per-image pricing at $0.03-$0.055, and open-weight variants you can self-host. Choose Midjourney for creative work, Flux for product pipelines.
Q: Does Midjourney v7 have an API? A: No public API exists as of August 2026. Enterprise conversations happen but there is no published pricing or self-serve access. For image generation as a backend service, use Flux, Ideogram, DALL-E, or Adobe Firefly APIs instead.
Q: Can I use Midjourney v7 images commercially? A: Yes on all paid tiers, with one caveat: companies with over $1M in annual revenue must be on Pro ($60/month) or higher. Solo creators, freelancers, and small studios are covered on Basic and Standard.
Q: How does Midjourney v7 video compare to Runway or Kling? A: Midjourney video is the best "same-aesthetic" option — the still and its motion come from the same model. But Runway Gen-4 and Kling 2.1 produce longer, sharper, more controllable clips. Use Midjourney for short stylized loops; use Runway or Kling for serious motion work.
Q: Does Midjourney v7 render text inside images properly? A: Not reliably. Short single words often work, but full sentences and complex typography still break. If your image needs legible text — posters, logos, UI mockups, book covers with titles — use Ideogram 3.0.
Q: What happened to Midjourney v8? A: Version 8.1 shipped April 30, 2026 as an incremental point release on the v7 architecture, adding Raw Mode, HD 2K output, and faster generation. It is available to all subscribers at no extra cost — the "v7" naming persists because the underlying model family has not changed.
Q: Is Midjourney worth $30/month if I already pay for ChatGPT Plus? A: Yes if you ship visual work regularly. ChatGPT Plus's GPT Image is fine for one-off illustrations, but Midjourney's aesthetic default, Draft Mode, Style Reference, and Personalization produce better results per hour of your time. If you generate images only occasionally for slides, ChatGPT Plus is enough.