Midjourney V8.1 became the default model on midjourney.com on June 10, 2026, after a roughly six-week alpha period that began with the V8.0 Alpha on March 17 and continued through the V8.1 Alpha that arrived on April 14 and reached release on April 30. The default cutover happened quietly. Most users opening the web app over the past two weeks have been generating with V8.1 without explicitly opting in. The improvements are real and substantial: HD output by default at meaningfully higher speed, dramatically better text rendering, and the restoration of the signature aesthetic that the V8.0 Alpha had drawn community criticism for losing. The caveats are real too: some features (Omni Reference, character reference) still silently fall back to V7 because the V8-native versions are not finished, and the broader competitive landscape for AI image generation has continued to shift around Midjourney in ways that affect how this version should be evaluated.
This piece walks through what V8.1 actually changes versus the V7 era that preceded it, the V8.0 Alpha to V8.1 restoration story that explains why this release feels different from a typical version bump, the parts of the workflow that still depend on the prior generation, the pricing and platform situation, the competitive context in mid-2026, the business situation including the active Disney and Warner Bros. Discovery lawsuits, and when V8.1 is the right tool for a given creative workflow. The goal is the full picture in one read for anyone evaluating whether to upgrade their Midjourney workflow or to pick a Midjourney alternative.
The short version is that V8.1 is the strongest Midjourney has been in many months. It is the right default for nearly every use case that was already a good Midjourney fit. It is not the right answer for use cases where competing platforms now lead (photorealism through FLUX.2 Pro, text-heavy typography through Ideogram 3, the absolute capability ceiling through OpenAI’s GPT Image 2 and Google’s Imagen 4). Midjourney’s strongest remaining position is concept art, moodboards, atmospheric exploration, and the distinct stylistic register that the platform has been refining since v3 and that V8.1 restored after the alpha cycle nearly lost.
The version timeline
A short recap of how we got here is useful because the alpha cycle for V8 was unusual in its length and in the community response.
V7 launched April 3, 2025 and became the default on June 17, 2025. The release was well-received. V7 brought improved coherence, better hand rendering, and a meaningful step up on multi-subject scenes. The "Midjourney look" was present and the community treated it as the strongest Midjourney to date.
V8.0 Alpha shipped to alpha.midjourney.com on March 17, 2026 as the first preview of the next generation. The alpha was opt-in and limited. The technical changes were substantial: faster generation, better text rendering, native HD. But the aesthetic shifted in a direction the community did not like. Outputs read as flatter, more over-polished, less distinctive. The widely-shared community reaction was variations on "they killed V7’s soul." The criticism was loud enough that Midjourney addressed it directly in the V8.1 release notes.
V8.1 Alpha shipped April 14, 2026 with explicit work to restore the signature Midjourney aesthetic that V8.0 Alpha had lost. The release notes called this out: the team had pulled in image prompts, image weights, the prompt shortener, an updated Describe model, and more stable moodboards and style references, all aimed at making V8.1 feel like Midjourney while keeping the V8.0 technical improvements.
V8.1 reached general release on April 30, 2026 on midjourney.com. It became the default on June 10, 2026. As of this writing (late June 2026), V8.1 is what new prompts use unless the user explicitly selects an older version, and the V8.0 path is largely deprecated in favor of V8.1 for any user who was not specifically attached to the V8.0 Alpha behavior.
The V8.1 release also introduced Draft mode (lower-cost fast iteration with a more constrained model) and the --preview flag that lets users opt into early model features before they ship to the main version. Both are operational quality-of-life improvements that have been well-received.
What V8.1 brings versus V7
The capability changes from V7 to V8.1 fall into four categories.
Speed. Standard generations are approximately 5x faster than the V7 equivalent. HD mode (the new default for high-resolution output) is 3x faster and 3x cheaper than the V7 HD path that required separate upscaling. Standard mode is 50 percent faster and 25 percent cheaper than V7 standard. The combined effect is that the iteration cycle for a typical Midjourney session is meaningfully tighter than the V7 era. Working through a series of variations on a prompt that would have taken half an hour now takes ten minutes.
Native HD. V7 produced lower-resolution output by default with HD requiring an explicit upscale step. V8.1 produces 2K HD natively without upscaling. The HD output has higher fidelity than the V7 upscale produced and arrives in a single step rather than two. For workflows that needed HD (almost all professional creative work), this is a meaningful pipeline simplification.
Text rendering. This is the most-discussed V8 capability change. V7’s text rendering was famously poor: words in prompts would come out as garbled approximations of letters that looked vaguely like text but were not legible. V8.1 renders text dramatically better. Words placed in quotes in the prompt come out as legible signs, labels, posters, and similar text-bearing elements. The improvement is large enough that workflows that previously had to composite text into Midjourney output as a separate step (in Photoshop, in a design tool) can now ask Midjourney to render the text directly. The text rendering is not yet at the level of Ideogram 3 (which is the current leader in this specific capability) but is competitive with everything else.
Prompt adherence. Complex multi-element prompts (a specific color palette plus a specific spatial arrangement plus specific lighting plus specific materials) come out closer to what the prompt described than V7’s output would have. The improvement is most visible on prompts with multiple specific constraints; simple prompts were already handled well by V7 and the improvement is less obvious there.
The V8.1 release also restored several features that the V8.0 Alpha had broken or removed: image prompts (using an existing image as a prompt input), image weights (controlling how much an image prompt influences the output), the prompt shortener (an automatic rewrite of long prompts into the shorter form V8 responds to better), and an updated Describe model (taking an image and producing the prompt that would generate it). The restoration of these features is what made V8.1 a release the community could accept; without them, V8 would have been a regression for many established workflows.
The honest caveat: Omni Reference and character reference
V8.1 has two specific feature gaps that users should know about before assuming V8.1 is the full picture.
Omni Reference (--oref) is Midjourney’s multi-subject reference feature, used for putting specific characters, objects, or settings into a generation. As of V8.1, Omni Reference is not yet native to V8. When a user adds the --oref flag to a V8.1 prompt, the system silently falls back to V7 to process the generation. The output uses the V7 model rather than V8.1.
Character reference (--cref) has the same situation. The V7-era --cref flag is not yet supported natively in V8.1 and falls back to V7 when used.
Midjourney has acknowledged this in the documentation and has indicated that V8-native versions of these features are in development. No specific timeline has been published. Users whose workflows depend heavily on character consistency or multi-subject reference will continue to get V7 quality on those specific operations even though the rest of their session is V8.1.
The practical consequence is that V8.1 is the right default for most workflows but not a full V7 replacement for the specific reference-dependent workflows. The honest read is to use V8.1 for the bulk of generation work and keep V7 in mind for the specific cases where reference fidelity matters most.
Pricing and tiers
The pricing structure is unchanged from the V7 era:
- Basic at $10/month: 3.3 fast hours
- Standard at $30/month: 15 fast hours, relax mode included
- Pro at $60/month: 30 fast hours, stealth mode included
- Mega at $120/month: 60 fast hours
Annual pricing offers a 20 percent discount. Extra fast hours are $4/hour. Companies with over $1 million in annual revenue must be on Pro or Mega tier per Midjourney’s commercial-use terms. The pricing tiers are unchanged in absolute terms even though V8.1 generates faster and cheaper per image, which means each tier effectively delivers more value than the equivalent V7-era tier. There is still no free tier; Midjourney has been consistent on this point.
The platform surfaces
The web app at midjourney.com is now the primary platform. The Discord bot that was the original Midjourney interface still works for users who prefer it, but Midjourney has been steering new users to the web app for several quarters and the web app’s feature surface is more complete. The Discord experience is increasingly maintained-as-legacy rather than actively developed.
The mobile experience is a responsive web app rather than a native app. The browsing and gallery experience works on mobile; the prompt-and-generation flow is workable but is designed primarily for desktop use.
Midjourney TV, the showcase surface that lets users browse other users’ generations, is still operational. It is more of a community feature than a creative tool.
The official API shipped as a closed beta in late 2025 and remains in limited-access closed beta as of mid-2026. The API has not opened to general availability and access is largely restricted to specific partners. Third-party wrappers (UseAPI, ImagineAPI, and others) continue to fill the gap for developers who need programmatic access, with the caveat that those wrappers are unsupported by Midjourney and operate at the user’s own risk.
Video generation shipped in June 2025 with the V1 video model. The feature is image-to-video, producing 5-second clips that can be extended to 21 seconds. The video output is competent for short looping clips and for adding subtle motion to still concepts; it is not at the level of Runway Gen-4, Pika 2, or Sora 2 for narrative video work.
The competitive landscape
The AI image generation market in mid-2026 has fragmented by use case, and Midjourney’s position has shifted accordingly. The current state:
For raw capability ceiling, the arena leaderboards rank OpenAI’s GPT Image 2 at #1 and Riverflow 2.0 Pro at #2. Midjourney V8.1 is competitive but not dominant on these leaderboards.
For photorealism in production pipelines, FLUX.2 Pro from Black Forest Labs has become the default choice for many workflows. The combination of strong photoreal output and a more accessible API has shifted a substantial fraction of professional photorealism work to FLUX.
For overall quality plus text rendering combined, Google’s Imagen 4 has emerged as the strongest balanced option. Imagen 4’s integration into Google’s broader workflow tools (Workspace, Slides, the broader Gemini surface) gives it natural adoption in those environments.
For typography-heavy work, Ideogram 3 remains the specialized leader. Anyone whose work depends on tight integration of legible text and design elements still reaches for Ideogram for those specific use cases.
For vector output and brand systems, Recraft owns the specialized lane. Recraft V4.1 in particular has become the standard for AI-generated SVG and brand-system work.
Stable Diffusion / Stable Cascade and its successors continue to serve the self-hosted and customization-heavy use cases. The open-weights ecosystem around SD has continued to be vibrant, with derivative models and LoRA fine-tunes covering specialized aesthetics that the major commercial platforms do not target.
Midjourney V8.1’s strongest position in this fragmented market is concept art, moodboards, atmospheric and stylized work, and the distinctive aesthetic register that the platform has refined over many versions. The aesthetic restoration in V8.1 is the specific reason this version matters; without it, the V8 path would have diluted Midjourney’s strongest differentiator.
The honest reading is that Midjourney is no longer the universal "best AI image generator" in any single capability dimension. It is the right answer for specific use cases and a competitive option for many others, but the universal-best framing that fit the V5/V6 era no longer fits the multi-vendor fragmented market of 2026.
The business situation
Midjourney’s business situation in mid-2026 includes several active threads worth knowing about.
Revenue. David Holz has publicly said the company surpassed $200 million in annual recurring revenue and has continued growing since. Midjourney remains profitable and self-funded.
Funding posture. Still no venture capital. No acquisition. No IPO. Midjourney has been the most prominent example of a profitable AI company that has refused outside funding, and that stance has been maintained through the V8 release cycle.
Headcount. Approximately 40 to 45 employees as of mid-2026, which is small for a company with the platform’s user base and revenue. The lean operation is part of the explanation for both Midjourney’s release cadence (faster than competitors with much larger teams) and for the specific feature gaps that take time to close (Omni Reference and --cref for V8, the API for general availability).
Litigation. Two active major lawsuits as of mid-2026:
- Disney and Universal v. Midjourney (case 2:25-cv-05275, filed June 2025) is still active. The post-mediation status conference is scheduled for August 31, 2026, with motions due November 23, 2026. The parties are in private settlement conference. No settlement has been reached.
- Warner Bros. Discovery filed a separate copyright suit on September 4, 2025.
Both cases involve allegations that Midjourney’s training data and output infringe the studios’ copyrighted material. Midjourney’s defense argument follows the broader generative-AI fair-use line that other AI companies are also litigating. The eventual outcomes will affect Midjourney specifically and the broader generative-AI category generally.
The Getty Images suit (against Stability AI, not Midjourney) was largely dismissed on jurisdictional grounds.
Partnerships. Midjourney signed a licensing deal with Meta on August 22, 2025 for the aesthetic technology underlying Midjourney’s distinctive look. The deal terms have not been publicly disclosed in detail. The partnership signals that Midjourney’s aesthetic IP has commercial value beyond the consumer subscription product.
Midjourney Medical. In June 2026, Midjourney announced a Midjourney Medical division. Details are thin in public information. The division appears to focus on medical imaging applications using Midjourney’s underlying model capability, though the specific products and customer story are not yet documented publicly.
When V8.1 is the right tool
The honest decision framework for Midjourney V8.1 in a creative or design workflow:
Use V8.1 as the default for any work that fits Midjourney’s existing strengths: concept art, moodboards, atmospheric imagery, stylized illustration, ideation, and the distinct aesthetic register Midjourney specializes in. V8.1 is the strongest version for these use cases in the platform’s history.
Use V7 specifically for workflows that depend on Omni Reference or --cref for character or multi-subject consistency. The V8 fallback to V7 makes this transparent; the user does not need to manually downgrade.
Use a competing platform for use cases where Midjourney is not the strongest answer:
- Production photorealism: FLUX.2 Pro
- Heavy text or typography: Ideogram 3
- Vector or brand systems: Recraft V4.1
- Overall capability ceiling for varied work: GPT Image 2 or Imagen 4
Use multiple platforms for serious design work that spans multiple use cases. The fragmented market means most professional creative workflows benefit from access to two or three platforms rather than committing entirely to one.
The decision is not "Midjourney or something else." It is "where does Midjourney fit in a multi-platform creative toolkit." V8.1 makes Midjourney a stronger participant in that toolkit but does not reclaim a universal-default position the platform has not held since 2024.
Frequently asked questions
Will my V7 prompts work in V8.1? Mostly yes. The basic prompt syntax is unchanged. Some specific parameters (style codes, image weights) may produce slightly different output in V8.1 than in V7 because the underlying model is different. Test before committing a V7-era workflow to V8.1 for production work.
Can I still use V7 if I prefer it? Yes. The version selector in the web app lets you specify which version to use for a generation. V7 will remain available for at least the foreseeable future.
Has video generation improved in the V8 cycle? The V1 video model is the only video model Midjourney has shipped; V8.x is the image model line. There has been no V2 video model announcement. The video capability is unchanged from the June 2025 V1 release.
Is the Midjourney API available yet? The closed-beta API exists but general availability has not been announced. Most developers are still using third-party wrappers, with the caveats that those are unsupported by Midjourney and may stop working without notice.
How does V8.1 compare to GPT Image 2 specifically? GPT Image 2 ranks higher on the current arena leaderboards for raw capability across mixed tasks. Midjourney V8.1 retains the edge on the specific aesthetic style work that the platform is known for. For a designer choosing one to use, the answer depends on which kind of work dominates.
Will the Disney lawsuit affect users of Midjourney? Probably not directly in the near term. The litigation is about the company’s training data and output, not about end users. The eventual outcome could affect what kinds of prompts produce results that are commercially safe to use, which would matter for users in copyright-sensitive workflows.
Is there a free trial for V8.1? No. Midjourney has never had a free trial and has been consistent on this point. The $10/month Basic tier is the lowest entry point.
Does V8.1 support custom training or LoRAs? No. Midjourney does not support custom training in the way Stable Diffusion does. The style codes and the Describe feature are the closest things to customization, and they operate on top of the base model rather than modifying it.
What is Midjourney Medical? Announced June 2026 as a new division applying Midjourney’s model technology to medical imaging applications. Specific products and customer-facing details have not been publicly disclosed beyond the announcement.
Should I upgrade from V7 to V8.1? If your workflow doesn’t depend on Omni Reference or --cref, yes. The speed, HD, and text-rendering improvements are substantial and the V8.1 release restored enough of the V7 aesthetic that the upgrade is mostly improvement. If your workflow does depend on those reference features, stay on V7 for now and let V8.1 handle everything else, since V8.1 automatically falls back to V7 for those operations anyway.
