On November 10, 2026, at Adobe MAX 2026 in Miami Beach, Florida, Adobe announced new features for Firefly, its AI service for creatives. The announcement adds a feature for editing images and other assets together, tools for building reusable production workflows, and support for prototyping websites and apps from natural-language instructions. Adobe's aim is to let Firefly handle not only generating images and video, but also the editing, review and workflow management that follow.
The new features fall into three distinct uses. "Firefly Design" is an environment for combining images and other assets, then editing and reviewing the deliverables. "Firefly Graph" connects multiple processing steps to build production workflows that can be reused. "Firefly Builder" creates prototypes of websites and apps you can actually interact with, guided by natural-language instructions.
Adobe had already announced a limited beta on June 18 of the "AI Assistant," which carries out multi-step tasks from natural-language instructions, along with an integrated production environment linking generation to editing. This is therefore not the first time conversational controls have been introduced. The November announcement adds a mechanism for reviewing the plan before the AI begins work, a model selection feature, and more concrete capabilities such as editing deliverables, reusing production workflows and creating prototypes.
Firefly Design Brings Asset Editing and Review Into One Place
Firefly Design is a production environment that handles vector and raster images, motion assets such as animations, and even 3D and print materials on a single canvas. It is designed so that users can not only place AI-generated assets, but also edit them, adjust layouts and carry out reviews in the same environment. The goal is to manage workflows that combine assets in multiple formats entirely within Firefly.
PSD, AI and SVG files can be placed as linked assets. For example, if you update the original file in Photoshop, the change is reflected where it has been placed. The review screen accepts comments and approvals, and finished deliverables can be exported in formats such as SVG, MP4 and print-ready PDF. The system reduces the need to revise assets and review deliverables in separate places, connecting the flow from editing to approval.
Image and video editing features have also been expanded. Firefly Image gains new or enhanced tools for partial edits, including "Edit Object" for editing objects, "Edit Text" for editing text, "Relight" for adjusting lighting, and "Retouch" for correcting images. Firefly Video can handle story structure, scenes and shots, and lets users assemble clips on a simple timeline. It also supports adding subtitles and generating scene transitions and supplementary footage, according to Adobe.
The preset "Quick Actions" make it easy to run commonly used operations such as removing objects, generating thumbnails and creating product ads. However, these are features as described by Adobe; the quality of the editing results, and whether areas outside the specified part of an image are fully preserved, have not been confirmed by independent testing. Before adopting them in real production, users will need to verify them against their own assets and quality requirements.
Users also get more control over how AI agents operate. "Plan Mode" lets users review, modify and approve the AI's plan before it runs. Because automatic execution can also be selected, human approval is not mandatory for every operation. "Model Selector" lets users either specify which AI model to use within a session or leave the choice to automatic selection by the system.
There is also a feature for setting rules on formatting, color schemes and branding and applying them consistently across multiple projects. "Firefly Boards" gains a feature for developing ideas and concepts through dialogue with the AI. Adobe also introduced an always-on agent that keeps working while the user is away, organizing assets, checking quality and preparing materials for each medium, but it is still in preview.
Firefly Graph Reuses Workflows, While Builder Prototypes Working Apps
"Firefly Graph" deals not with deliverables such as images themselves, but with the processing steps used to create them. Users place AI models, Creative Cloud functions, input assets and more as processing units called "nodes," and connect them to build a series of production steps. Results from one step are passed to the next, and users can check intermediate outputs while changing inputs and connections.
A workflow, once built, can be saved and reused, and outputs from multiple AI models can be compared. The built-in AI Assistant helps build and optimize workflows. Whereas the conventional assistant takes a description of the desired result in natural language and handles the processing, Graph is distinguished by letting users visually assemble the flow of processing itself, then edit and reuse it later. For example, when applying several editing steps repeatedly to the same type of asset, a saved workflow can be used rather than rebuilding the operations each time.
"Firefly Builder," meanwhile, is a web app that creates website and app prototypes from natural-language instructions and reference assets. It keeps the look consistent according to preset design rules, and users can revise it by selecting on-screen elements directly or by talking to the AI. The screens it creates are not static design images but prototypes that can be clicked and used, and can be shared as such, making them suitable for checking screen transitions and usability before development begins.
Once a prototype has been reviewed, a full set of HTML code can be handed to developers or web development tools for further work. While Graph saves production workflows that can be run repeatedly, Builder outputs an interactive prototype and its code. However, a Builder prototype alone does not cover developing and operating an app that runs in a production environment. Publishing a real service requires accounting for the development work that follows once the output code is handed over.
Availability and Conditions to Check Before Adoption
Based on Adobe's November 10, 2026 announcement and product FAQs, we have organized the new features by use, availability and points to note when adopting them. Individual image edits, deliverable production, repeatable workflow building and creating working prototypes each rely on different features at different stages of availability. The table below summarizes information as of the announcement and does not compare the performance or quality of each feature.
| Feature | Main use | Availability | Conditions to check before adoption |
|---|---|---|---|
| Edit Text, Retouch, Relight, Quick Actions | Partial edits to text and images, standard processing | Generally available from November 10, 2026 | Whether the editing features and preset operations fit your production goals |
| Firefly Design | Editing and reviewing deliverables that combine multiple assets | Public beta | Available to all users during the beta. Details of usage tiers and pricing after general availability are unconfirmed |
| Firefly Graph | Building and reusing workflows by connecting nodes | Public beta | Nodes that use generative models consume generative credits. The number required varies by model and node and is shown before running |
| Firefly Builder | Creating interactive prototypes and exporting HTML code | Limited beta via waitlist | English only. Does not support exporting a full set of code for mobile apps |
Quickly fixing part of an image and building a production workflow that combines multiple processes call for different features. Firefly Design is available to all users during the public beta, but usage tiers are planned for general availability. Pricing after general availability and the details of what each plan includes could not be confirmed as of the announcement.
Those adopting Firefly Graph for ongoing production work should also pay attention to generative credit consumption. Not every node consumes credits, but nodes that use generative AI models do, and the amount varies with the model and the processing. The credits required can be checked on each node before running, so when the same workflow will be run many times, it is important to understand the consumption of each step. Evaluation should also account not only for the number of generations but for the revision work needed to reach the desired result. At this point, no independent measurements have been presented of how much Graph shortens working time or reduces costs.
Firefly Builder is available from desktop and mobile browsers, but currently supports English only and cannot export a full set of code for mobile apps. And even when a clickable prototype can be created, implementation and operation afterward require separate development work. Before adoption, organizations should also settle who will review and approve prototypes and which development environment will receive the exported HTML code.
Adobe also stresses that work created in Firefly can be moved into Creative Cloud applications for further finishing. In addition, it revealed a vision for a local MCP that would let external AI agents edit native files on a PC while preserving editable structures such as layers. However, the limited beta for that is planned for early 2027, so it is not available alongside this announcement.
The announcement shows Adobe trying to expand Firefly from a simple generative AI tool into a production environment covering editing, workflow management and prototyping. To judge the real benefits of adoption, though, organizations will need to look beyond generation speed to Graph's credit consumption, revision work, and the handoff from Builder to developers. Whether the next person in line can take over the deliverables and workflows created, and use them repeatedly in real work, will be the key to judging their practical value going forward.