Midjourney V8.1, ChatGPT Images 2.0, and Adobe Firefly: Everything Happening in AI Image Generation in 2026

The AI image generation ecosystem at its most active

The first half of 2026 has been remarkably intense for AI image generation tools. Midjourney launched its most ambitious model to date, OpenAI redefined what it means to create images inside a chat interface, Adobe transformed Firefly into a full-fledged creative automation platform, and the legal debate over Hollywood copyrights reached a turning point. For entrepreneurs and marketing teams across Latin America, understanding these developments is key to making smarter decisions about which tools to adopt and how to use them.

Midjourney V8.1: the fastest and sharpest model in its history

Midjourney made V8.1 the default model for all users in June 2026, replacing V7, which had held that position since early in the year. The leap forward is significant: according to the platform's official documentation, V8.1 is the fastest model in Midjourney's history, with standard generations completing in just 4 seconds and HD mode finishing in 12 seconds.

The journey here began on March 17, 2026, when V8 Alpha launched — built on a completely rewritten codebase and considered the most important advancement since V5. Then, on April 14, 2026, V8.1 arrived with concrete improvements that transformed the user experience: HD mode is now three times faster and three times cheaper than before, standard resolution is 50% faster and 25% more affordable, and a Run as HD button was added to convert any standard image to high definition in a single click.

V8.1 also generates images natively at 2K resolution and significantly improves text rendering within images — historically one of the weakest points across AI generators. The platform also introduced a style exploration mode with the --sref random command, which generates 24 images in different styles in draft mode to streamline creative iteration.

Another standout feature is video generation from still images: users can convert a static image into a 5-second clip and extend it up to 21 seconds. The platform also improved its conversational voice mode, which now accesses image references and user settings during the session.

In the background, the community is already feeding ranking data to help shape improvements for V8.2, the next version currently in development.

OpenAI replaces DALL-E with ChatGPT Images 2.0

On April 21, 2026, OpenAI unveiled ChatGPT Images 2.0 (known internally as gpt-image-2), the direct successor to GPT Image 1.5, which launched in December 2025. This release marks the end of the DALL-E era as the primary image generation engine inside ChatGPT: back in March 2025, DALL-E 3 had already been replaced by GPT's native image generation capabilities, and now Images 2.0 takes that concept even further.

The most important new feature in this model is the introduction of Thinking mode, which integrates the reasoning capabilities of OpenAI's O-series into visual generation. This means that before creating an image, the model researches, plans, and reasons through the visual structure. According to the specialized outlet Infobae, when this mode is active, Images 2.0 can search the web in real time, generate multiple distinct images from a single prompt, and review its own outputs before delivering them.

The model operates in two versions:

• Instant Mode: available for free to all ChatGPT users. Generates images quickly for everyday tasks.
• Thinking Mode: exclusive to paid subscribers (Plus, Pro, and Business). Enables prior reasoning, real-time web search, and generation of sequences of up to 8 visually coherent images.

On the technical side, ChatGPT Images 2.0 reaches a maximum resolution of 2K and addresses one of the biggest long-standing challenges in generative AI: producing legible text within images. The model also adds support for non-Latin scripts such as Japanese, Chinese, Hindi, and Korean — a meaningful upgrade for teams with global audiences. According to OpenAI's own data, the image generation feature in ChatGPT now surpasses 1 billion images generated per week.

For developers, API access is priced at $8 per million input tokens and $30 per million output tokens — 20% cheaper than the previous version.

Adobe Firefly: from image generator to creative automation platform

Adobe wasn't standing still either. In an update announced on June 18, 2026, Firefly moved beyond being a standalone image and video generator to become a creative automation layer that coordinates multi-step tasks through a unified conversational interface. The shift is structural: the original Firefly worked by generating an image from a prompt and then requiring manual export to other apps. Now, the workflow is continuous.

Among the most relevant Firefly updates in 2026:

• Custom models in public beta (since April 2026): brands can train their own AI models using their own images, capturing visual style, color palette, lighting, and character traits for consistent reproduction.
• Over 30 integrated models, including Google's Nano Banana Pro, OpenAI models, and Runway, all accessible from Firefly's model selector.
• Quick Cut: turns raw footage into a structured first edit in seconds.
• Unlimited generation of images and video on select plans.
• Project Moonlight (in private beta): an agentic interface that connects Photoshop, Express, and Acrobat to execute real actions across tools based on natural language instructions.

In addition, Firefly's image editor has offered features since March 2026 such as Generative Fill, Generative Remove, Generative Expand, and Generative Scale — all accessible directly from the web and mobile interface.

The unexpected pivot: Midjourney enters the world of medicine

The most surprising news of the month came on June 17 and 18, 2026, when Midjourney announced its first hardware project: Midjourney Medical, a new division focused on diagnostic imaging. Its first product is the Midjourney Scanner, a full-body ultrasonic tomography device that promises to scan the human body in approximately 60 seconds, with no radiation and no magnetic fields.

The system works by lowering the user onto a platform inside a water tank, where approximately 500,000 ultrasonic sensors capture data from multiple angles to build three-dimensional images of muscles, fat, bones, and organs. The technology is based on an exclusive licensing agreement with Butterfly Network, signed in November 2025 and valued at up to $74 million over five years.

Midjourney CEO David Holz described the device as the first new method of full-body imaging in decades. The company plans to open its first location — called Midjourney Spa — in San Francisco by late 2027, with the ambitious goal of deploying 50,000 scanners worldwide by 2031. However, the current prototype does not yet have FDA approval for diagnostic use, and the company will initially offer only body composition maps while pursuing regulatory clearances in stages.

The move drew skepticism from the medical community: some radiologists pointed to physical limitations of ultrasound when it comes to penetrating bone and deep tissue. Even so, the announcement had an immediate market impact: shares of Butterfly Network jumped 33% following the news.

The legal battle that will define the future of AI copyright

As the technology races forward, the legal front shows no signs of letting up. In early July 2026, Midjourney filed a motion in federal court seeking to compel Disney, Universal, and Warner Bros. to disclose how they use artificial intelligence in their internal processes. The dispute stems from copyright infringement lawsuits the three studios filed against Midjourney last year, alleging that the platform allows users to generate images of protected characters — such as Bart Simpson and Darth Vader — without authorization.

Midjourney is defending its position under the fair use doctrine, arguing that the studios themselves may be using AI in similar ways within their own production pipelines. The company is seeking access to AI business plans, training datasets, model weights, and executive presentations on artificial intelligence. In June, a magistrate judge limited that access to consumer-facing AI applications only — a ruling Midjourney is currently appealing.

The outcome of this case could set a foundational precedent on transparency in AI model training and the applicability of fair use in the sector, with direct implications for the entire generative AI industry worldwide.

What this means for entrepreneurs and SMBs in Latin America

The global generative AI market was valued at $53.7 billion in 2025 and is expected to reach $83.3 billion in 2026, according to Global Market Insights. Image generation is one of the highest-adoption segments in advertising, e-commerce, and product design. For Latin American SMBs, this translates into a real opportunity: according to Utilia AI data, more than 54% of small and medium-sized businesses across the Americas already use AI in some form, and nearly 49% use generative AI in their operations.

Competition among platforms is driving prices down and quality up. Midjourney V8.1 offers plans starting at $10 per month. ChatGPT Images 2.0 in Instant mode is free. Adobe Firefly is included with Creative Cloud. Choosing the right tool is no longer just about image quality — it's about workflow, legal compliance, and the real cost per generated image.

In this context, platforms like Doobl.IA give marketing teams across Latin America a way to go even further: integrating visual content generation with automated social media publishing through AI agents, cutting production time and maintaining brand consistency — without needing large teams.