Midjourney is a subscription image and video generation service from Midjourney, Inc., and its versions, plan limits and ownership terms decide what a team can produce and who owns it. Midjourney's documentation lists model versions from V6 (released 20 December 2023) to V8.1 (released 30 April 2026), a separate Niji line for anime and illustrative styles, and a video model that animates still images into clips of up to 21 seconds. It is sold through four monthly or annual plans that buy GPU time rather than a set number of images, and its Terms of Service set a USD 1,000,000 revenue line above which companies must hold a Pro or Mega plan to own their outputs. This page sets out versions, plans, video limits, licensing, privacy and moderation facts from Midjourney's own pages. It is written for designers, marketing teams and the legal, privacy and security staff who approve creative tools.
Last checked: 2026-11-15.
This is an independent reference page from NHI Mgmt Group based on public provider documentation; it is not sponsored or endorsed by any provider. It draws on Midjourney's public documentation as of November 2026. Product features change, so confirm settings against Midjourney's current documentation.
Key takeaways
- V8.1 is the newest version. Midjourney's models page lists V8.1, released 30 April 2026, and a Midjourney update post announces that V8.1 replaced V7 as the default.
- Plans buy GPU time. Basic, Standard, Pro and Mega cost USD 10, 30, 60 and 120 a month and include 3.3, 15, 30 and 60 Fast GPU hours; annual billing is 20% lower.
- Revenue decides the plan. Companies, or their employees, with more than USD 1,000,000 annual revenue must use Pro or Mega to own their outputs under the Terms of Service.
- Outputs are public by default. Content is public and remixable unless a Pro or Mega subscriber uses Stealth Mode, and Midjourney holds a perpetual licence to content and outputs.
- Video is image-to-video. Clips start at 5 seconds and extend to a maximum of 21 seconds, at 480p (SD) or 720p (HD on Standard and above).
What is Midjourney?
Midjourney is a generative AI service that turns text prompts, and optional image references, into images and short videos. Users work through the Midjourney website or its Discord server; the V8 alpha ran on a separate site, alpha.midjourney.com. Midjourney does not offer a public developer API, and its community guidelines prohibit scripts, third-party apps, "unauthorized APIs" and reselling access, with one account per person for individual use. That makes it a creator tool rather than a platform to build products on.
The service is organised around versioned models selected with a version setting or parameter. Features such as Omni Reference, Draft Mode, upscalers and the Editor are tied to particular versions, so the version a team chooses changes what it can do.
Key facts
| Field | Value (as of 2026-11-15) |
|---|---|
| Provider | Midjourney, Inc. |
| Modality | Image generation and editing; image-to-video |
| Latest version | V8.1, released 30 April 2026 |
| Anime and illustration line | Niji 7, launched 9 January 2026, developed with Spellbrush |
| Access | Midjourney website and Discord; no public developer API |
| Plans | Basic, Standard, Pro, Mega (USD 10 to 120 a month) |
| Free plan | Not stated by Midjourney on its plans page |
| Maximum image size stated | HD images at 2048px on V8.1 |
| Maximum video length | 21 seconds (5 seconds plus four 4-second extensions) |
| Open weights | Not stated by Midjourney; service is subscription only |
| Provenance signal (C2PA or watermark) | Not stated by Midjourney on the pages checked |
| Terms of Service effective date | 27 May 2026 |
| Privacy Policy last updated | 2 June 2025 |
Models and versions
| Version | Released | Default period | Notes from Midjourney |
|---|---|---|---|
| V6 | 20 December 2023 | 14 February to 30 July 2024 | Character Reference and multi-prompting supported |
| V6.1 | 30 July 2024 | Until 16 June 2025 | About 25% faster than V6; its upscalers, Pan, Zoom Out and Editor are reused by later versions |
| V7 | 3 April 2025 | From 17 June 2025 | Omni Reference, Draft Mode and Conversational Mode; V7 Omni Reference stays available while a V8 version is trained |
| V8.0 Alpha | 17 March 2026 | Not default | Alpha site only, Fast mode only, limited time; update post said it would be deprecated two weeks after V8.1 became default |
| V8.1 | 30 April 2026 | Announced as default, replacing V7 | "Fastest model so far"; native HD at 2048px; requires an unlocked global V7/V8 personalization profile |
As of 2026-11-15. Midjourney's models page states that "the current default Midjourney version is V7", while its update post "V8.1 is now the default model" announces that the default moved from V7 to V8.1. The pages were not in step on the check date, so confirm the default in account settings before relying on either. Midjourney's update posts on the V8 releases carry no publication dates, so the dates above come from the models page.
What changed in V8?
- Speed: the V8 alpha post described generation as about five times faster; the default post gives about 4 seconds for SD and 12 seconds for HD on V8.1.
- Resolution: a native HD mode renders at 2K; Midjourney says V8.1 HD images are twice the size and four times the resolution of V7 images. HD costs 1.33 minutes of GPU time per job on the models page.
- Text rendering: Midjourney says text renders better when the wanted words are placed in quotation marks.
- Continuity: V8 accepts V7 personalization profiles, moodboards and style references, so a team's look carries over.
- Tooling: V8.1 added "Run as HD" to rerun a job at HD, brought back image prompts and image weights, and added a Prompt Shortener and a longer Describe output.
Which features work on which version?
| Feature (models page chart) | V6 | V7 | V8.1 |
|---|---|---|---|
| Maximum aspect ratio | 14:1 | 14:1 | 14:1 (4:1 for HD) |
| HD images (2048px) | No | No | Yes |
| Upscalers | Subtle and Creative | Uses V6.1 upscalers | No |
| Character Reference | Yes | No | No |
| Omni Reference | No | Yes | No |
| Style Reference, personalization, moodboards | Yes | Yes | Yes |
| Quality parameter | 0.5, 1, 2 | 1, 2, 4 | No |
| Draft Mode | No | Yes | No |
| Conversational Mode | No | Yes | Yes |
| Turbo Mode | Yes | Yes | No |
| Relax and Fast modes | Yes | Yes | Yes |
| Seed | Yes | Yes | Yes (99% identical) |
Pricing and plans
| Plan (as of 2026-11-15) | Monthly | Annual | Fast GPU time per month | Relax Mode | Video | Stealth Mode |
|---|---|---|---|---|---|---|
| Basic | USD 10 | USD 96 (USD 8 a month) | 3.3 hours (200 minutes) | Not included | SD | Not included |
| Standard | USD 30 | USD 288 (USD 24 a month) | 15 hours | Unlimited images | SD and HD | Not included |
| Pro | USD 60 | USD 576 (USD 48 a month) | 30 hours | Unlimited images and SD video | SD and HD | Included |
| Mega | USD 120 | USD 1,152 (USD 96 a month) | 60 hours | Unlimited images and SD video | SD and HD | Included |
Source: Midjourney plans page, checked 2026-11-15. Extra Fast GPU time costs USD 4 an hour on every plan. Plans renew automatically each month or year until cancelled, and annual plans are paid upfront. The Terms of Service say subscribers are not refunded for the current subscription period but are not charged after it ends.
- Concurrency: Basic runs 3 fast image jobs; Standard 3 fast or relax; Pro and Mega 12 fast or 3 relax. For video, Basic runs 1 fast job, Standard 3, Pro 6 fast or 3 relax, and Mega 12 fast or 3 relax.
- Queues: up to 10 queued jobs on every plan; maximum repeat or permutation size of 4 (Basic), 10 (Standard) and 40 (Pro and Mega).
- Cost multipliers: in the V8 alpha, HD, quality 4, style reference and moodboard jobs were four times slower and cost four times as much "for now"; the V8.1 post then said HD became three times faster and three times cheaper, and standard resolution 50% faster and 25% cheaper.
Who owns Midjourney images?
Ownership is set by Section 4 of the Terms of Service (effective 27 May 2026), and the plans page applies the General Commercial Terms to every plan.
- Ownership: users own their outputs ("Assets") "to the fullest extent possible under applicable law", subject to the agreement and third-party rights. Upscaling someone else's image does not transfer ownership of it.
- Revenue threshold: a company, or an employee of a company, with more than USD 1,000,000 in annual revenue must hold a Pro or Mega plan "to own Your Assets". Procurement should size the plan to the organisation, not the individual seat.
- Licence to Midjourney: users grant Midjourney "a perpetual, worldwide, non-exclusive, sublicensable no-charge, royalty-free, irrevocable copyright license" to their content and outputs, and it survives termination.
- Visibility: content is public and remixable by default. With Stealth Mode on Pro or Mega, Midjourney agrees to make "best efforts" not to publish outputs, which does not cover work made in shared spaces such as Discord channels.
- Other people's rights: users may not infringe intellectual property or repost others' creations without permission, and for the Editor and video they must have permission from anyone depicted in uploaded or generated content.
- Takedowns: rights holders send notices to [email protected]; repeat infringers can lose access, and users can file counter-notices.
- Disputes: California law, with binding arbitration in Santa Clara County after a 30-day negotiation period.
Copyright protection for AI-generated images differs by country, and owning an output under a contract is not the same as holding copyright in it. See the AI Training Data Provenance and Copyright Exposure Guide for that question.
Features
Midjourney's V1 video model animates a starting image. The image can be a Midjourney image or an upload, with an optional text prompt for the motion. Midjourney's announcement describes an "automatic" setting that writes a motion prompt and a "manual" setting where the user describes the movement, plus low motion (still camera, ambient scenes) and high motion (subject and camera both move).
| Video setting (as of 2026-11-15) | Value from Midjourney's video page |
|---|---|
| Starting length | 5 seconds |
| Extensions | 4 seconds each, up to four, maximum 21 seconds; each extension costs the same as an initial generation |
| Resolution | SD 480p; HD 720p on Standard, Pro and Mega, Fast Mode only |
| GPU cost at batch size 4 | SD 8 minutes; HD 26 minutes (batch 2 and 1 cost about a half and a quarter) |
| End frame | Set with --end and an image URL |
| Looping | --loop or the Loop checkbox reuses the start frame as the end frame |
| Relax Mode | Pro and Mega, SD only |
| Blocked jobs | Not charged; GPU time is deducted only when results are shown |
At launch, Midjourney said a video job cost about eight times an image job, produced four 5-second videos, and was web-only, and that prices would be adjusted over the first month.
Availability
Midjourney is used through the Midjourney website or its Discord server, on one of four paid plans (Basic, Standard, Pro and Mega). The V8 alpha ran on a separate site, alpha.midjourney.com. Stealth Mode, which keeps images and videos private, is only available on the Pro and Mega plans, and companies with more than USD 1,000,000 in gross annual revenue must buy Pro or Mega.
Provenance and labelling
Midjourney's plans, models, video, terms and community pages checked on 2026-11-15 do not state whether outputs carry C2PA Content Credentials or an invisible watermark. Not stated by Midjourney is not the same as absent: teams that need provenance should inspect exported files with a C2PA verifier and keep originals. Midjourney's community guidelines do ask users to be transparent about the source of generated images, and disclosure duties such as Article 50 of the EU AI Act apply to the deployer whatever the tool does. See the AI Content Provenance and C2PA Guide and the EU AI Act Article 50 Transparency Obligations Guide.
Data and privacy
| Topic (Privacy Policy, last updated 2 June 2025) | What Midjourney states |
|---|---|
| Data collected | Username, prompts (text, images and content from spoken input), uploaded photos, videos and documents, IP address, usage data, cookies, contact details |
| Payment data | Handled by third-party processors; Midjourney keeps confirmation of payment |
| Training use | The policy lists data collected through training Midjourney machine learning algorithms as a category; it does not set out an opt-out for prompts or uploads |
| Retention | Only as long as necessary for the stated purposes, legal compliance and disputes |
| User rights | Access, correction and deletion requests via midjourney.com/account; objection, restriction, portability and complaint to a regulator |
| International transfers | Standard contractual clauses for EEA, Swiss and UK data |
| Visibility | Outputs public by default; Stealth Mode on Pro and Mega |
Because outputs are public by default and uploads become part of the prompt record, teams should keep confidential material, unreleased product designs and personal data out of prompts on plans without Stealth Mode.
Where identity and security fit
Midjourney accounts sign in through the website or Discord, so the account is a personal identity holding the organisation's prompts, uploads and outputs. Treat it like any SaaS seat: require company-managed sign-in where available, enable multi-factor authentication on the underlying account, remove access when people leave, and record which plan each seat holds so the revenue rule is met. Shared logins break both the one-account rule and the audit trail.
Release history
- 20 December 2023: V6.
- 30 July 2024: V6.1, about 25% faster than V6.
- 3 April 2025: V7, default from 17 June 2025.
- 17 March 2026: V8.0 Alpha, on the alpha site only.
- 30 April 2026: V8.1, later announced as the default in place of V7.
Dates are from Midjourney's models page; its update posts on the V8 releases are undated.
Limits
- No developer API: automation and third-party apps are prohibited, so Midjourney cannot be built into pipelines the way API-based models can.
- Version feature gaps: V8.1 drops upscalers, Omni Reference, Draft Mode, Turbo Mode and the quality parameter listed for V7, and uses V6.1 for Pan, Zoom Out and the Editor.
- Short video: 21 seconds maximum and 720p at most.
- Documentation drift: the models page and update posts disagreed on the default version on the check date, and update posts are undated.
- Shared accounts: one account per person for individual use, so teams need a seat per user rather than a shared login.
What are Midjourney's content rules?
- Safe for work: the community guidelines prohibit gore, adult content, sexualised imagery and any content that sexualises minors, plus racist, homophobic or other derogatory content.
- Deception: no misinformation, political campaign imagery, election influence, or content meant to deceive or defraud; users are asked to be transparent about the nature and source of generated images.
- Real people: imagery of real people may not be used to harass, abuse, defame or harm, and Midjourney says it takes "especially strict action" against attempts to make sexualised images of real people in the Editor.
- Enforcement: some inputs are blocked automatically, but an unblocked prompt is not necessarily allowed; breaches can lead to a warning, a time-out or a ban, and any job can be reported with Report Job.
- Age: users confirm they are at least 13 and meet the digital consent age in their country.
For organisational rules on likeness, disclosure and approved tools, see AI Image and Video Generation at Work, and for misuse risks the Deepfake and AI Impersonation Guide.
Practitioner checklist
- Confirm the organisation's annual revenue and buy Pro or Mega for every seat if it exceeds USD 1,000,000.
- Turn on Stealth Mode for confidential work and keep that work out of shared Discord channels.
- Record the model version used for each campaign, and check the default in settings after each release.
- Keep prompts free of personal data, client names and unreleased designs on plans without Stealth Mode.
- Get written permission from anyone depicted before uploading or editing their image.
- Label Midjourney images as AI-generated where law or policy requires, without relying on embedded metadata.
- Give each user their own account, with multi-factor authentication, and remove it at offboarding.
- Track GPU hours and extra GPU purchases, since HD images and video draw time far faster than SD images.
Standards and references
- Midjourney: Model versions
- Midjourney: Plans
- Midjourney: Video
- Midjourney: Terms of Service
- Midjourney: Privacy Policy
- Midjourney: Community Guidelines
- Midjourney: V8.1 is now the default model
- Midjourney: V8 alpha
- Midjourney: V8.1 alpha
- Midjourney: Introducing our V1 video model
Related NHI Mgmt Group resources: AI Image and Video Generation at Work · AI Content Provenance and C2PA Guide · AI Training Data Provenance and Copyright Exposure Guide · Deepfake and AI Impersonation Guide