Adobe today detailed a major expansion of its Firefly generative AI capabilities, unveiling updated image and vector models, making its video generation tools widely available, and introducing a redesigned web application that, perhaps most notably, now hosts AI models from key competitors like Google and OpenAI.
Announced at the Adobe MAX London event, this evolution aims to unify Adobe’s AI-powered tools for image, video, audio, and vector generation into a single, cohesive platform. The strategy positions the Firefly web app as a more centralized hub for AI-assisted creativity, offering users broader choice while keeping them within Adobe’s ecosystem. A dedicated Firefly mobile app is also set to launch later this year.
Opening Firefly to Competitors
The move appears geared towards maintaining Creative Cloud’s central role for professionals, even as numerous standalone AI tools gain traction. By incorporating models such as Google’s Imagen 3 (image) and Veo 2 (video), OpenAI’s GPT image generation, and Black Forest Labs’ Flux 1.1 Pro directly within the Firefly interface, Adobe gives users access to different aesthetic styles and capabilities without leaving its environment.
Adobe confirmed that models from Runway, Pika, Luma, Ideogram, and fal.ai are expected in the coming months. “We know that customers have preferences,” explained Ely Greenfield, Adobe’s CTO of Digital Media, in comments reported by The Verge. “We wanted to make sure that they had choice.”
Usage of these integrated third-party models will be managed through Adobe’s existing generative credit system, a payment model first outlined in September 2023, though specific credit costs for competitor models haven’t been announced yet.
Adobe also requires that these partners adhere to its AI policy which prohibits training models on user content processed through Firefly, aiming to address privacy concerns often associated with third-party AI tools. How smoothly these different models perform within the Firefly wrapper, and whether the credit costs prove economical compared to direct subscriptions, remain points to watch.
Adobe’s Enhanced AI Models
Alongside the third-party integrations, Adobe launched its own upgraded models. Firefly Image Model 4, available today in the web app, and the more advanced Image Model 4 Ultra promise better prompt adherence and realism compared to predecessors like Firefly Image 3 (which hit public beta last July).
Adobe describes Image Model 4 as “Ideal for rapid ideation and everyday creative needs,” covering “90 percent of typical creative requirements quickly and inexpensively,” while Ultra “shines in rendering photorealistic scenes, human portraits, and small groups” for “highly complex needs, when precision and clarity are paramount.” Both can output images up to 2K resolution, but the Ultra version will consume more generative credits per image than the standard version.
The Firefly Video Model is also now generally available, moving beyond the limited beta phase first detailed last October. This model enables the creation of video clips up to five seconds long at 1080p resolution using text or image prompts within the Firefly web app.
Adobe states that for Firefly Premium plan members, access to the Firefly Video Model is unlimited. The model also powers the Generative Extend feature in Premiere Pro, which itself became generally available with 4K and vertical video support on April 3.
Generative Extend allows editors to add two seconds to video clips or ten seconds to ambient audio tracks, helping fix shots that are slightly too short without requiring reshoots. Furthermore, a Firefly Vector Model is now generally available, underpinning a Text to Vector feature for generating editable vector graphics like icons and patterns directly from prompts.
Platform Expansion and Collaboration Tools
The redesigned Firefly web application acts as the central interface for accessing both Adobe’s and the integrated third-party models, along with persistent access to users’ generation history. Within this updated web experience, Adobe introduced Firefly Boards in public beta.
This tool provides a collaborative canvas, described by some sources as resembling Figma but with embedded AI generation, designed for mood boarding, storyboarding, and shared brainstorming.
Looking beyond the desktop, Adobe confirmed a dedicated Firefly mobile app for iOS and Android will arrive “later this year,” offering features like Text to Image, Generative Fill, and Generate Video for on-the-go creation.
For developers, Adobe’s Text-to-Image API and Avatar API are now generally available via Firefly Services, with a new Text-to-Video API available in beta.
Content Credentials and Training Transparency
Coinciding with the Firefly updates, Adobe launched the public beta of its free Adobe Content Authenticity web application today. This tool allows creators to attach Content Credentials – metadata detailing how content was made, who made it, AI usage (and which model, including third-party ones via Firefly), and training opt-out preferences – to images (supporting batch application), regardless of the creation tool used.
This standard, accessible via ContentCredentials.org, aims to bring more transparency to digital media provenance. Adobe reiterated that its own Firefly models adhere to its AI ethics principles and are trained on licensed datasets like Adobe Stock and public domain content.
“Our commitment ensures that every Firefly output is not only visually stunning but also IP-friendly and commercially safe for professional use,” the company stated on its blog. Adobe reported that over 22 billion assets have now been generated using Firefly tools since their initial rollout, up significantly from the 8 billion reported last May.
Last Updated on May 8, 2025 10:10 am CEST
