Adobe Firefly adds reusable characters and creative memory | Magica
Adobe adds reusable subjects and memory to Firefly’s campaign workflow
Editorial Team
••📖7 min read
Adobe’s Firefly update saves recurring visual subjects and creative preferences, extending an existing push into repeatable production. Memory has narrower reach than the assistant, while campaign demonstrations leave reliability and savings unmeasured.
Firefly can save characters, objects and locations as reusable elements for later projects.
Creative memory is a separate feature, enabled by default and rolling out in stages within the assistant and Firefly Boards.
Adobe’s campaign demonstrations show the intended workflow; they do not establish its reliability or cost savings in production.
Adobe, the creative software company behind Photoshop, Premiere and Illustrator, is adding reusable visual subjects and creative memory to Firefly, its studio for AI generation and editing. Announced at its Adobe MAX conference on November 10, the additions address a practical campaign problem: carrying the same character, product and creative requirements through multiple deliverables and revisions.
Adobe’s illustration of its visual-story workflow connects reusable character elements with successive shots. Image: Adobe. Source: Adobe Community.
Adobe’s element-creation guide shows the reference views used to define a recurring character. The interface’s consistency label is not a measured production success rate. Image: Adobe. Source: Adobe Help Center.
A reusable subject for successive shots
Firefly has grown from an image generator into a studio combining Adobe tools and partner AI models across media. Adobe’s announcement presents the Adobe Creative Agent, which powers the beta Firefly AI Assistant, as the coordinator. Its proposed advantage is continuity between tasks: users can set formats, colors, brand guidelines, characters and products for reuse across projects, instead of treating each generation as an isolated request.
The practical mechanism for a recurring subject is an element. According to Adobe’s , users save an uploaded or generated image as a character, location or object, then reference it with an @ mention in a chat. A new chat requires another initial @ reference; within the same conversation, ordinary language can refer back to it.
Binance Wallet says subscriptions for its Anthropic IPO xStocks campaign reached the 500 million USDC limit. Final allocations, token delivery and applicable refunds remain pending; the cap does not establish proceeds received by the AI company.
An element contains more than a name attached to a picture. Adobe’s creation guide describes reference images, identifying descriptions and additional generated views. A character can have front-face, front-body and back-body references; an object can have front and back views. Users can edit the description, including attributes they want the generation to ignore.
That gives a creator a reusable visual specification instead of requiring a fresh upload and description for each request. It also introduces material to inspect: generated views become part of the subject definition. Adobe says these references help maintain appearance and identity; the documentation does not provide a measured consistency rate.
A report from the conference describes a demonstration by Elise Swopes, a surrealist artist and Adobe’s senior evangelist and community relationship manager. She supplied an approved fragrance pitch and reference assets, then asked the assistant to create elements of herself and a perfume bottle. It used them in still images and video scenes. After she supplied four photographs of herself wearing blue, the software revised the character element and every shot using it.
Swopes said that change would otherwise have required a reshoot. Her role explains both her practical familiarity with image making and her stake in the demonstration: this was an Adobe showcase, not an independent production test.
Adobe’s feature walkthrough explicitly connects reusable elements to a visual-story workflow, with clips generated together or shot by shot. The intended unit of work is a sequence of related deliverables.
Adobe’s Memory artwork illustrates a separate feature for retaining creative preferences. Adobe says Memory is enabled by default and is rolling out in stages. Image: Adobe. Source: Adobe Help Center.
Memory has different rules
Saving a character is distinct from asking the assistant to remember how its user works. Adobe’s memory FAQ says the assistant retains lasting creative preferences, such as colors, image sizes, file types and whether to ask before creating something. Users can explicitly request a memory, or the assistant can infer one from conversation.
The distinction between preferences and memories affects control. Manually set preferences are seen with every message. Memories are consulted when deemed relevant, with only a few relevant ones used. A format that must always apply therefore belongs in preferences, according to Adobe’s guidance. Instructions in the current conversation take priority over a memory.
Memory is on by default. When the assistant saves something on its own, it may not notify the user each time. Users can inspect, change or delete memories in settings or through conversation. Switching memory off stops new saving and its use in new chats, but preserves existing memories. Deleting a chat does not delete the memories extracted from it.
The saved wording is not fixed either. Adobe says changing one memory can cause the assistant to reword related memories. For a production brief, reviewing what was retained is a separate task from approving an individual image.
There is also a narrower deployment boundary than the broad studio pitch might suggest: the FAQ says memory works in Firefly AI Assistant and Firefly Boards, its workspace for mood boards, storyboards and creative concepts. It is rolling out in stages and is not yet in other Adobe apps. The assistant’s presence in a workspace does not by itself establish that memory works there.
Adobe’s Firefly Graph illustration depicts connected inputs and processing steps for a reusable workflow. Graph is offered in beta. Image: Adobe. Source: Adobe Help Center.
Adobe is selling coordination across tools
This is an expansion of an existing workflow strategy. Adobe’s release history lists the conversational assistant’s beta introduction in April 2026, already able to plan and execute multistep work. That month also brought reusable production pipelines and brand validation capabilities. November adds more ways to carry context into the work rather than introducing repeatability from scratch.
Adobe combines its editing tools with models from providers including Google, Kling and OpenAI in the same environment. Its new model selector lets users choose a model or delegate selection to the agent; Plan Mode lets them review, modify and approve the proposed work. Firefly Graph, also in beta, connects models and Adobe tools into reusable workflows. The update therefore offers several ways to repeat work, with elements and memory supplying different kinds of context.
Those controls matter as Adobe moves toward recurring automation. A separate report on a company preview describes the assistant proposing footage organization, color processing, audio cleanup and backups for user approval. Adobe labels its always-on agents a preview. The report places the move alongside proactive agents from Anthropic and OpenAI; it does not establish equivalent creative capabilities.
Competition also extends to the underlying editing software. The same report identifies Artcraft as a free Adobe alternative made possible by coding models from Anthropic and OpenAI. Asked about it, Adobe defended its products on professionalism, security and reliability. That is Adobe’s competitive claim, rather than a demonstrated comparison of the two products.
Adobe is also testing connections that let external chatbots operate its desktop creative apps. Coordination need not begin inside Firefly: retaining a role for Adobe’s editing tools is another route through this market.
The next test is the revision bill
Adobe says Edit Text, Retouch, Relight and Quick Actions are generally available; other new capabilities are in beta, with always-on agents still presented as a preview. A creator evaluating reusable characters should distinguish those stages before building a recurring production process around them.
Adobe’s launch announcement links to pricing but does not specify a separate charge or usage allowance for elements and memory. The sources reviewed here supply no measured reduction in rejected generations, manual corrections, production time or campaign cost from those features. A subscription price alone would not establish the cost of an accepted campaign asset.
The useful test is whether a saved subject remains recognizable across a real sequence, whether a brief change reaches the intended assets, and how much correction and regeneration the result requires. Adobe has shown a mechanism for carrying context forward. The outstanding evidence is whether that mechanism makes the finished campaign cheaper and more dependable.
Tencent says US restrictions on remote computing will likely slow its next AI model while leaving cloud revenue largely protected. Southeast Asian operators face licensing and financing uncertainty, though US cloud access and domestic demand offer partial alternatives.
Microsoft is rolling out Agent 365 Business Analytics through its Frontier early access program, pairing AI spending with estimates of assisted work. The dashboard expands its existing analytics as agent workloads add usage charges, but its dollar-value estimates do not establish financial returns.
Microsoft’s Ignite upgrades package enterprise knowledge, skills and tools into reusable business plugins. The expansion builds on IQ’s June release and competes with existing graph retrieval services; total savings and portability remain unproven.
The European Central Bank warns that debt and interlinked financing could carry an AI investment reversal into credit markets and the European economy. Resilient earnings, limited direct bank lending and stronger bank buffers temper its comparison with past bubbles.
Reported introductions helped Alice Huang secure Alibaba as a chip-rental customer and obtain financing. The relationship raises questions about Nvidia's overseas sales network, but does not itself establish illegal exports.
KPMG says AI agents pursued some issues more deeply in a parallel UK audit, though humans still had to correct their work. Client fee savings remain unconfirmed, even as KPMG International has negotiated lower fees from its own auditor by citing AI.
Payhawk announced a $280 million all-equity Series C led by Growth Equity at Goldman Sachs Alternatives, valuing it at $2 billion. The funding backs an expansion from managing business spending into broader AI finance operations, where competitors already combine automation, payments and controls.
A 16-country survey of 15- to 22-year-olds finds a median of 33% have used AI as a companion or friend. Users have higher adjusted odds of loneliness, but the research cannot establish whether AI helps, harms or replaces human connection.
South Korean President Lee Jae Myung has ordered a review of rewards of at least 100 million won for reporting technology-assisted exam cheating and legislation requiring technology companies to supply evidence. Ahead of the November 19 college entrance exam, the proposals would add informant incentives and post-exam investigations to classroom checks, but implementation rules remain unsettled.
Microsoft is pitching Copilot Workspaces as a shared environment for people and AI agents. Team collaboration is already in early access, but personal Autopilot integration is still coming soon, with final agent pricing unannounced.
Google is rolling out file references and sharing for Gemini skills as adult personal accounts lose the ability to use Gems. Automatic conversion preserves supported material, but file limits and missing tools can interrupt existing workflows.
Muddy Waters disclosed a short position in Nvidia while questioning a possible connection involving senior adviser Ajay K. Puri. Its own account leaves the key identity question unanswered, separating its new claims from an existing server-diversion prosecution.
Nvidia reported $115.8 billion in quarterly revenue in its first earnings release since co-founder Jensen Huang resigned. Strong demand gives interim CEO Colette Kress a financial cushion, while a campaign for Huang’s return collides with unresolved investigations and the board’s reported opposition.
OpenAI launched GPT-6.2 Astra and Sol for professional work, cutting Astra’s standard input and output token prices by 20%. The models offer cheaper paths through selected tasks, but research evaluations, reporting failures and unfinished independent safeguard testing complicate the case for broader autonomy.
Microsoft’s first live speech-to-speech model offers OpenAI API compatibility and planned enterprise integrations. Its published evaluations show a low-cost option, with several rivals completing more agentic tasks.
Canberra-based Trellis Data is offering its 220-billion-parameter Agentaus 1 model to business and government. Its pitch goes beyond keeping data in Australia, but performance, deployment costs and a planned capital raising remain key tests.
Perplexity is adding direct collaboration between frontier models to Computer, letting a lead agent coordinate up to 16 supporting agents. Company-reported benchmarks suggest mixed teams can outperform repeated copies of one model, but leave customer costs and production reliability unresolved.
Marsh reported 105 client cyber incidents in the last two weeks of October, against 48 in the final two weeks of May. The broker links the increase partly to AI, while warning that insurance panel access may not guarantee an available response team.