Midjourney has released V8.1, a refinement of its image generation model that addresses one of the biggest frustrations users faced with V8: unpredictable quality swings. The new version, which became available on the alpha website in mid-April 2026, was developed over three to four weeks based on direct user feedback and promises to deliver more reliable results across the board .

What Changed in Midjourney V8.1?

The core improvement centers on consistency. V8 had what the Midjourney team describes as a "high highs, low lows" problem, where one generation might produce stunning results while the next attempt yielded something rough or unfinished. V8.1 smooths out these inconsistencies, delivering more predictable quality across generations .

Beyond consistency, V8.1 brings practical improvements that affect daily workflows. Generation speeds are faster, high-resolution outputs look sharper, and the model handles detail and blur with better precision. The team is also experimenting with different generation modes, similar to a draft mode but with more granular control over quality and speed tradeoffs .

Which Features Are Still Missing?

Not every requested feature made it into V8.1. Three notable gaps remain: negative prompting (the "--no" command), the "--tile" function, and OREF (object reference), which allows users to reference specific objects from one image in another. OREF is among the most requested features, and the team confirmed that the underlying data and approach are ready, but extensive testing is required before release. No firm timeline has been announced .

The team is prioritizing quick wins first while building larger systems in parallel. Near-term improvements will focus on image sharpness, detail handling, blur reduction, and upscaling quality. The most transformative feature on the horizon, however, is the editing model.

How Will the Editing Model Change Creator Workflows?

The editing model represents the most significant workflow shift coming to Midjourney. Currently, if an image is 90 percent correct, users must regenerate from scratch and hope for better results. An editing model would let creators fix what's wrong while preserving what works, eliminating the regeneration gamble .

An 8.0-based version of the editing tool already exists internally, and the team is upgrading it to work with V8.1. The expected timeline is a few weeks, making this one of the most anticipated near-term releases. For creators working at scale, this feature could cut iteration time dramatically.

Steps to Maximize V8.1 Results Today

• Use Style References: The team specifically encouraged leveraging style references to close any quality gaps and achieve more consistent aesthetic results across multiple generations.
• Apply Personalization Codes: Custom personalization codes help fine-tune outputs to match your specific creative vision and reduce unwanted variation.
• Refine Prompt Tuning: Detailed prompt engineering remains one of the most effective ways to guide V8.1 toward your intended output, especially as the model becomes more consistent.

What's Coming After V8.1?

V8.0 was always intended as an alpha release and won't remain available long-term. The team plans to move forward with a closed alpha, open beta, and stable release cycle, keeping fewer legacy models available. Some of V8.0's edge-case strengths will be absorbed into V9 rather than maintained as separate options .

V8.2 will likely bring improved datasets and refinements, but V9 is where ambitions expand significantly. The next major version will focus on better coherence, improved text rendering, stronger style control, and multi-image referencing beyond just faces. Larger datasets are now available, which opens the door to more complex input and output combinations .

V8.1 isn't yet the default model due to server constraints and infrastructure transitions. Some staffing limitations are also slowing deployment speed. Broader availability is expected within one to two weeks, and the team considers it ready to serve as the default once infrastructure is fully transitioned and OREF is implemented .

Why Midjourney Is Staying Focused on Images for Now

Despite the success of video generation tools in the broader AI market, Midjourney is not pursuing video generation at this moment. The team noted that training video models is expensive, and the path to profitability remains unclear. They mentioned plans for another video model eventually, with long-term interest in real-time simulations rather than traditional linear video production .

This strategic focus on image generation allows the team to iterate rapidly on core features like editing, consistency, and style control. For creators relying on Midjourney for visual content, V8.1 represents a maturation of the platform, moving from experimental tool to reliable production system. The editing model coming in weeks could be the feature that finally makes Midjourney feel less like a slot machine and more like a true creative partner.