Midjourney v7

Midjourney v7 is the seventh-generation text-to-image generation model from Midjourney, the self-funded San Francisco image-generation lab founded by David Holz, released in 2025 and available through the Midjourney Discord bot, the Midjourney web app, and a limited API beta. The model generates high-resolution images from text prompts and image references, with documented strengths in aesthetic quality, stylistic range, and photorealistic rendering, including features for character consistency and style transfer introduced with the v7 generation. As of April 2026, Midjourney v7 is broadly recognized as the leading image-generation system for creative and professional aesthetic output, supported by a subscriber community that generates approximately $500 million in annual revenue without external venture capital.

At a glance

• Lab: Midjourney
• Released: 2025 (v7 initial release; became default model June 2025)
• Modality: Image (text-to-image generation)
• Open weights: No (closed)
• Output resolution: Up to 2:1 or 1:2 aspect ratios at high resolution; default outputs up to approximately 1024 pixels on the longer side, with upscaling available to 2048 pixels
• Pricing: Subscription tiers: Basic ($10/month, ~200 fast GPU minutes), Standard ($30/month, 15 fast GPU hours), Pro ($60/month, 30 fast GPU hours, stealth mode), Mega ($120/month, 60 fast GPU hours). Annual billing available at approximately 20% discount. Subscriber counts and tier pricing subject to revision at https://www.midjourney.com/account
• Distribution channels: Midjourney Discord bot (https://discord.gg/midjourney), Midjourney web app (https://www.midjourney.com), API access (limited beta)

Origins

The Midjourney product launched in open beta in July 2022, distributed exclusively through a Discord bot. Users joined the Midjourney Discord server, submitted text prompts through bot commands, and received generated images in shared channels alongside thousands of other users. The community-first distribution model was architecturally unusual for a commercial AI product: rather than a private prompt interface, early generation happened publicly, which created a discovery dynamic that accelerated both community formation and the development of prompt-engineering culture among early adopters.

David Holz founded Midjourney in 2021, drawing on his background as co-founder of Leap Motion, the gesture-input hardware company where he served as CTO until its 2019 acquisition. Holz framed the new lab as a research organization focused on expanding creative possibility through AI, rather than a conventional commercial startup. The structuring decision to operate without external venture capital, which has been sustained through April 2026, reflects this orientation.

The version history from v1 through v7 represents a substantial capability trajectory over approximately three years. V1 through v3 (2022) established the Discord-bot distribution model and early stylistic identity. V4 (2022) was a significant quality step with more coherent compositions and better anatomy, expanding Midjourney's reputation among professional designers. V5 and v5.2 (2023) introduced photorealistic rendering capability and user-controlled stylization parameters. V6 and v6.1 (2024) improved prompt fidelity and moved to natural-language prompting, allowing users to write in ordinary sentences rather than comma-separated keyword lists. V7 (2025) brought photorealistic improvements, more precise text-and-image prompt handling, richer textures, and more coherent details; v7 became the default model in June 2025.

The Niji model, a sibling product developed in collaboration with Spellbrush, provides an anime-optimized variant of the Midjourney image generation capability. Niji 6 is the current generation and maintains a separate user base particularly active in anime, illustration, and adjacent creative communities.

Midjourney reached profitability within approximately one year of its July 2022 launch, reporting revenue of approximately $200 million in 2023, $300 million in 2024, and $500 million in 2025 through subscription fees alone. The company operated with a notably small team relative to its revenue, with team size reported in the dozens through 2023 and into 2024. A hardware division was established in August 2024 with the hiring of a former Apple Vision Pro engineer to lead it; the hardware effort's capital intensity has generated speculation that Midjourney may consider external funding for the first time, though no external capital round had been publicly announced as of April 2026.

Capabilities

Midjourney v7's principal strength is aesthetic image quality across a wide range of styles. The model produces outputs that professional designers, illustrators, and marketing creatives have consistently preferred in informal comparisons, with a distinctive visual coherence that has been difficult for competing systems to replicate. This quality is most evident in stylized, artistic, and compositionally complex prompts where the output is evaluated on visual impact rather than strict prompt adherence.

The v7 generation introduced several capability improvements over v6.1:

Photorealism. V7 improved naturalistic rendering of people, environments, and materials. Lighting, skin texture, and material properties are rendered with greater detail than in prior versions.

Prompt fidelity. A longstanding criticism of Midjourney relative to DALL-E 3 and Imagen was that Midjourney deprioritized strict prompt adherence in favor of aesthetic interpretation. V7 improved fidelity on multi-element descriptions with specific spatial or compositional requirements.

Personalization. Linked to a user's prior generation history, this feature adjusts v7's outputs to align with a user's expressed aesthetic preferences based on past ratings and selections, creating a form of model customization without explicit per-generation style prompts.

Style Reference (--sref) and Character Reference (--cref). --sref accepts one or more reference images and applies their visual style to a new generation. --cref applies character appearance from a reference image to a new scene, enabling consistent character depiction across multiple generations. These features address a recurring limitation of earlier diffusion-based systems: maintaining consistent style or character identity across a series of images.

Draft Mode and Omni Reference. Draft Mode produces reduced-quality outputs for rapid prompt iteration before committing to full-quality generation. Omni Reference (--oref) allows a single reference image to influence both style and character simultaneously.

The Niji 6 model, a sibling product developed in collaboration with Spellbrush, provides anime-optimized outputs on the same generation infrastructure, with its own distinct subscriber community.

Benchmarks and standing

Image-generation benchmarking is substantially less standardized than text-model benchmarking. There is no widely adopted composite leaderboard equivalent to the Artificial Analysis Intelligence Index. Evaluations typically combine human-preference side-by-side comparisons, FID (Frechet Inception Distance) scores, and capability-specific tests covering prompt adherence, text rendering, and photorealism ratings.

On the LMArena image arena, Midjourney v7 consistently appears in the top tier on aesthetic quality and stylized-output categories. It leads on creative, artistic, and visually complex prompts evaluated on aesthetic impact, and trails Imagen 4 Ultra and FLUX.2 on prompt-fidelity tasks for complex multi-element descriptions and text-within-image rendering legibility.

The Hugging Face Text-to-Image Leaderboard reflects the same pattern: competitive on quality dimensions overall, more variable on strict fidelity tasks.

Qualitative dimensions where Midjourney v7 leads: creative aesthetic output, stylistic range across photorealistic and artistic modes, photographic quality in portraits and environments, and professional-designer adoption. Qualitative dimensions where it trails or is comparable: text rendering within images (where DALL-E 3, FLUX.2, Imagen 4, and Ideogram lead), strict multi-element prompt fidelity, and agentic or workflow integration pending full API availability.

Benchmark leadership in image generation is point-in-time and prompt-category-dependent. Methodologies are not standardized and a new model release can shift the leaderboard substantially.

Access and pricing

Midjourney's primary distribution channels are the Discord bot and the web app.

The Midjourney Discord server remains the original and heavily used distribution channel. Users join the server at https://discord.gg/midjourney and submit prompts using the /imagine command and related bot commands. Generations appear in shared channels, creating a community discovery dynamic that has sustained one of the largest Discord communities globally.

The Midjourney web app at https://www.midjourney.com provides a browser-based interface with gallery views, generation history, and a conventional prompt interface. The web app launched in 2023 and has expanded progressively, with Personalization and reference-image tools available through it.

Subscription tiers as of early 2026: - Basic: $10/month or approximately $8/month annually. Approximately 200 fast GPU minutes per month. Suitable for low-volume personal use. - Standard: $30/month or approximately $24/month annually. 15 fast GPU hours per month, plus unlimited slow-mode generation at lower queue priority. The most common tier for regular creative use. - Pro: $60/month or approximately $48/month annually. 30 fast GPU hours, stealth mode (generations not visible in the public community gallery). - Mega: $120/month or approximately $96/month annually. 60 fast GPU hours. For high-volume professional use.

Annual billing reduces monthly cost by approximately 20% across all tiers. Additional GPU time can be purchased as top-up credits when monthly allocations are exhausted.

The API beta is available to qualifying subscribers and enterprise customers. Midjourney has moved cautiously from Discord-only distribution to the web app to the API, and as of early 2026, the API remains in limited access and is not publicly available to all subscribers. Developer programs and enterprise agreements can be inquired about through https://www.midjourney.com.

Comparison

The direct peer set for Midjourney v7 in April 2026 is the leading text-to-image generation systems:

• DALL-E 3 (OpenAI). The image-generation model with the largest consumer user base, distributed through ChatGPT and Microsoft Bing Image Creator. DALL-E 3 leads on text-within-image rendering and on free-tier accessibility; Midjourney v7 leads on aesthetic quality in most side-by-side comparisons. Midjourney's subscriber base self-selects for engaged creative users in a way that ChatGPT's broad audience does not.
• Imagen 4 (Google DeepMind). Google's fourth-generation image-generation model, available through Vertex AI and the Gemini app. Imagen 4 Ultra leads on photorealism and text-within-image rendering; Midjourney leads on stylized and creative-aesthetic outputs. Imagen 4 is structurally positioned for enterprise Google Cloud use; Midjourney is positioned for individual subscribers and creative professionals.
• FLUX.2 (Black Forest Labs). The leading open-weights-derived image-generation system from the original Stable Diffusion team. FLUX.2 leads on prompt fidelity for complex multi-element descriptions and on text rendering; Midjourney leads on aesthetic and stylistic creative output. FLUX.2's open-weights variants make it the preferred choice for developers building image generation into applications.
• Stable Diffusion 3.5 (Stability AI). Stability AI's current open-weights flagship. Generally trails FLUX.2 and Midjourney v7 on composite quality scores, but widely used in the open-source ecosystem for self-hosted and commercial applications.

Midjourney's distinctive position across this peer set combines creative-aesthetic leadership as the preferred tool of professional designers and illustrators, a subscriber-community moat built on the Discord-first distribution model, and the self-funded profitable business structure without external venture capital. Midjourney's users are highly engaged, generate community content that functions as informal marketing, and have stronger switching friction than casual users of ChatGPT image generation.

Outlook

Several open questions shape Midjourney's trajectory through 2026 and into 2027:

• V8 timeline and capability profile. The Midjourney version cadence has ranged from months to over a year between major releases. V8 is anticipated by the community, with open questions around whether it will close the text-rendering gap to competitors, further improve prompt fidelity, and maintain aesthetic quality leadership against advances from FLUX.2 and Imagen.
• API expansion strategy. Midjourney has moved deliberately from Discord-only to web to limited API. Whether the API opens broadly to subscribers and developers, and on what pricing model, will determine whether Midjourney becomes a platform for third-party applications or remains primarily a direct subscriber product. The API expansion is a significant strategic fork.
• The hardware effort. Midjourney's hardware division, established in August 2024, has not publicly released a product as of April 2026. Whether the hardware initiative leads to a launched product, and whether it requires external funding that would alter the company's self-funded structure, remains an open question.
• Competitive pressure on creative-aesthetic leadership. FLUX.2 and Imagen 4 are advancing on quality dimensions across all categories. The question of whether Midjourney's creative-aesthetic advantage is durable, or whether it narrows as other systems improve, is the central competitive uncertainty for the business.
• Training data and IP litigation. Midjourney has been named in litigation regarding AI training data and the use of copyrighted images in training. The outcome of ongoing legal proceedings, as well as regulatory developments in the EU and other jurisdictions around training data rights and synthetic image provenance, could affect the operating model.

Sources

• Midjourney official site. Product reference, subscription pricing, and access documentation.
• Wikipedia: Midjourney. Comprehensive company and product history covering the v1-through-v7 version lineage, founding, and commercial trajectory.
• The Information: Midjourney Revenue Now 'Significantly' Above $200 Million. Revenue reporting and commercial context.
• GetLatka: How Midjourney hit $500M revenue and 100K customers in 2025. 2025 commercial scale and subscriber data.
• Midjourney v7 release announcement via Midjourney Discord. Primary source for v7 capability announcements and feature documentation.
• DemandSage: Midjourney Statistics 2026. User and revenue statistics with community size data.
• Hugging Face Text-to-Image Leaderboard. Cross-model image-generation benchmark comparison.