Midjourney

by midjourney-company (San Francisco, led by David Holz)

Proprietary, subscription-only text-to-image (and short video) service, used via a web app and Discord.

See https://www.midjourney.com/

Current version: V8.2 (2026)

  • Midjourney’s own post on X (x.com/midjourney/status/2080781271043911807, quoted in search results: “We’re releasing V8.2 today and making it the default model… focused on aesthetics, personalization, and image quality”) and the docs “Version” page (search snippet; page itself returns 403) say V8.2 became the default on 2026-07-24, replacing V8.1; focus: aesthetics, image quality, personalization (larger image pool for personalization profiles, fewer random low-quality results).
  • Timeline (Wikipedia and scriptbyai.com timeline, secondary): V7 alpha 2025-04-03/04, default 2025-06-17; V8 alpha 2026-03-17 (faster, native 2K); V8.1 alpha 2026-04-14, main release 2026-04-30 (more V7-like aesthetic, cheaper HD), default 2026-06-11; V8.2 2026-07-24. Video V1 launched 2025-06-18 (Animate / image-to-video).
  • updates.midjourney.com shows a weekly “Alpha Changelog” through 2026-09-23 and a 2026-09-24 post “Edit updates, thumbnail previews, and more”: the web editor is changing quickly. Video is a separate system with its own parameters (--motion low|high, --loop, --end; --v 8.2 does not select a video model); the vendor pages fetched do not say whether a video model newer than V1 exists.
  • Examples that use --v 6 are historical. Use --v 8.2 (or the default) now; check the version docs for supported parameters.

Features

  • Text-to-image via /imagine (Discord) or the web app; four variants per prompt with upscale/vary/re-roll; web editor and edit tools.
  • Parameters: --ar (aspect), --stylize/--s, --seed, --no, --v (model version), --q; personalization profiles and omni-reference (third-party guides for V8.x).
  • Fast / Relax / Turbo modes and Stealth (private) mode on higher tiers; image blending, region editing, image-to-video.

Superpowers

Strongest at stylised, art-directed imagery: concept art, mood boards, thumbnails, visual exploration. Weaker than GPT Image (gpt-image-2-5) or Nano Banana (nano-banana-pro) for precise text rendering and instruction-following edits (general assessment, not benchmarked here).

Plans and licence

Subscription tiers per third-party 2026 guides (not the official page, which could not be fetched): Basic, Standard, Pro and Mega; Standard and up include unlimited Relax; Pro and Mega include Stealth; annual billing is discounted; no free trial. Current prices: see the vendor pricing page https://www.midjourney.com/account/subscribe. Paid-plan images are generally usable commercially; read the Terms of Service.

Disney and Universal sued in June 2025 and Warner Bros. Discovery in September 2025; the court consolidated the cases on 2025-11-04. Per Variety, The Art Newspaper and trackers (July 2026), the case is in discovery (fact discovery scheduled to close 2026-09-21; motions deadline 2026-11-23). No ruling on fair use and no settlement had been reported; a 2026-10-02 search shows the last known docket filing on 2026-09-23, expert discovery closing 2026-11-09, summary-judgment motions due 2026-11-23, mediation directed since late 2025, and 2027 as the earliest realistic trial; still no judgment or settlement (CourtListener docket 2:25-cv-05275, trackers).

Promptcraft

Subject, mood, lighting, lens, palette, style; use --no to exclude; save seeds; keep final assets outside the service; use Stealth for client work.

flux-dot-1, seedream-40, nano-banana, gpt-image-2-5, comfyui, midjourney-company

Limits

Midjourney’s docs and pricing pages returned 403, so the V8.2 default date is per Midjourney’s X post and docs snippet as quoted in search results (2026-10-02) plus Wikipedia; plan tiers are per third-party guides (albato.com, costbench.com, fluxnote.io).

Sources (accessed 2026-10-01 and 2026-10-02)