Accelerating Creativity and Productivity with Adobe Firefly

How is Generative AI evolving to empower creative professionals and redefine digital content creation at scale?

‍

Hosts Patrick Moorhead and Daniel Newman are joined by Adobe's Alexandru Costin, VP, Generative AI, for a conversation on accelerating creativity and productivity with Adobe Firefly. The session explores the transformative power of Gen AI, recent advancements in Adobe Firefly, and new opportunities and risks for creative workflows. Learn more about themes like the integration of AI into creative tools, maintaining commercially safe models, and embracing partner AI models within the Firefly app.

‍

Key Takeaways Include:

‍

🔹Product Innovation & Safety Commitment: An in-depth look at how Adobe Firefly, a family of commercially safe Generative AI models, is integrated into Adobe’s Creative Cloud apps and how its enterprise APIs and Custom Models facilitate scalable, secure content creation.

‍

🔹The Power of an Open Ecosystem: Insights into Adobe’s expansion with partner models from Google, OpenAI, and others, offering creators more flexibility while ensuring content provenance and reducing workflow disruptions with Content Credentials.

‍

🔹Quantifying Enterprise Impact: Real-world examples from organizations like Versuni and Amazon Fresh that are seeing significant reductions in asset creation time and a marked increase in creative efficiency, validating the tangible business impact of Gen AI.

‍

🔹The Future of Creative Workflows: A forward-looking perspective on Adobe's future plans for Firefly, including the development of AI agents designed to further empower creators and automate complex creative tasks, signaling the next wave of innovation.

‍

Learn more at Adobe

‍

Watch the full video at sixfivemedia.com, and be sure to subscribe to our YouTube channel, so you never miss an episode.

‍

Or listen to the audio here:

Disclaimer: Six Five Media is for information and entertainment purposes only. Over the course of this webcast, we may talk about companies that are publicly traded, and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors, and we ask that you do not treat us as such.

‍

Patrick Moorhead: Welcome back to Six Five podcast. We are talking about our favorite subject, for sure, Daniel. That is AI. But in this case, we are talking about a different kind of AI, a creative AI that can help even people like us who are not that creative, whether it's image, video, audio, ideation, whatever.

‍

Daniel Newman: Didn't you send me a text just recently about how important it is to be creative? And then you actually said to me, you are. That's you.

‍

Patrick Moorhead: I mean, there's business creativity, Daniel, and then there's audio, video, imagery.

‍

Daniel Newman: Well, I mean, it's. It's kind of the same. I mean, you see, you know, the great designs of. Of the evaluations we do, or you see the awesome template of 6 5, a lot of that great creativity. Remember that old picture of you and me? We had the goatees and we were a little bit bigger because we hadn't gone on these really great health journeys. Alexandra, about that. But in all seriousness, you're absolutely right. Look, we are in this moment right now, where AI is all the rage. Consumer business in our life and the government. It's the biggest media story every week. It's the biggest decision. Most of our foreign policy is being decided right now on who's going to lead and control. It is the, you know, the custodian of all of our data and what we want to give, what we don't want to give. But the truth is, we're also having a lot of fun with this stuff, Pat, like the things we can create. Someone sent me some images of me the other day. Hair. You know, I know we were doing some backstage stuff, and by the way, it wasn't really good hair. It was like a really high hairline. And I showed him to my wife, and I was like, in a suit. And basically they took a photo of me and they dressed me up, they gave me hair, they put some glasses on me. And this is a microcosm of this. Of this topic, though, Pat, is. We have entered a world where the democratization of creativity is accelerating and that people like you and me that maybe aren't able to extensively use the Adobe suite and know how to, you know, go deep on this stuff, can now do really awesome stuff with products like Firefly.

Patrick Moorhead: Yeah, that's right. And tools for consumers, tools for enterprises. Enterprises just have a. A completely different bar. And we're going to jump into that. We've talked about Adobe Firefly before, but let's get an update on where it's at and what is the incremental value to it, even when they're pulling in third party models. It is my pleasure to introduce Alexandru from Adobe. Great to see Alexandru. We talked a little in the green room. Not too sure how we met, where we met, but we've definitely met before.

‍

Alexandru Costin: Happy to be here.

‍

Daniel Newman: Yeah, thanks so much. Well, you heard my background and my backstory a bit here. And talking a little bit about what you guys are doing is deeply personal. And I mean the fact is that the technology you've created can be incredibly productive for businesses to move fast on things like design and doing so in ways that consider compliance and data and protection of trademarks and companies intellectual property. But at the same time we've also brought everything to the world in a way that just wasn't before. Like I want to design something cool, it's been brought and made possible. This is something that's deep, deeply entrenched in Adobe's ethos. You guys have been doing this as long as Adobe's been in business?

‍

Alexandru Costin: Been doing it for four decades. Yes, yes.

‍

Daniel Newman: And so this has to be kind of a moment where you're like, this is amazing. And this is also a moment of sort of introspection where every company is looking at their own products, their own services, their own. How do we continue to disrupt the way we did 40 years ago and do so into the future? So anyways, I'd love to just hear a little bit because I think Firefly was almost two years ago.

‍

Alexandru Costin: Now it's three years. It feels like a decade, but who's counting? Happy to talk about some of our.

Daniel Newman: What's kind of happened there, how it's advancing.

‍

Alexandru Costin: Yeah, yeah. So just to start quickly with a quick recap, we've been doing creativity tools at Adobe for a long time and we really pride ourselves in being in understanding the customer needs, helping them paint the image they have in their heads or the video they have, the story they want to tell, the emotion they want to transmit, use our tools to do that. And we've been doing it with computer vision with, we're doing it with AI actually for the last 10 years. Smaller models doing masking and segmentation, then smaller models called core gans that were doing some generation but not super high quality. But over the last three years we've seen these larger models that are generative AI capable not only of generating high quality content, but also deeply understanding user intent. That can be expressed in a prompt, can be spoken to, or can be controlled with novel ways of controlling. We launched Fireflies three years ago in 2023. And since then we've integrated it in our big flagship products like Photoshop, Illustrator, Premiere, in the new products like Adobe Express or the Firefly web application. And we've seen tremendous success. We've seen more than 28 billion images being generated or edited with Firefly. We've seen millions, tens of millions of videos being generated because we've expanded the capabilities of the Firefly family of models to support all the modalities, all the content types our customers were looking for, from images to videos to designs to vector, illustrations, all of those core capabilities now audio or entering the audio space, all of those core modalities and all of those core workflows in the content creation process. We've been thinking hard, both inventing the technology, the AI technology to shape, reshape those workflows, but at the same time, innovate in the way we integrate the GenAI capabilities into the actual tools that we have, tools that our customers are spending their basic lives in. It's an amazing journey and I'm happy to talk to you more about where we are today, some of the developments we're making and opening the ecosystems to partner models that are coming in, and we're offering our customers choice and some of the ideas we have for taking this innovation further.

‍

Patrick Moorhead: Yeah, I appreciate you catching us all up here. I get this question a lot. And it seems that only the largest companies on the planet seem to be creating their own models with these gigantic hyperscaler Data centers melting GPUs as I've seen a few people talk through. You actually have your own generative AI models, right? So you are in that class, the elite class of people who are doing this. Can you talk about why you're doing that, why it matters, why it's more than just, hey, this is really cool tech.

‍

Alexandru Costin: Yeah, for sure. From the beginning, when we think about the impact of AI on software, we're a very pragmatic company and we usually use open source technology, if it exists. Like, we're not shy of being pragmatic of serving our customer needs with what exists out there. But when the Genai wave happened, we talked to our customers, we talked both in the individual artists' space, freelancers, small businesses and enterprises, and we asked all of them, what are the characteristics of GenAI technology that they would use? And they all told us, gave us feedback that shaped the principles that also got us to build our own models first. They really wanted to make sure the Genai technology they want to use is not trained on their assets without their consent. Especially our customers that are the creatives, the people creating most of the content on planet Earth, they were very sensitive to this subject. They were also asking us to make sure that the technology we're building is not going to put them in a hot seat in terms of involuntary creating copyright infringement, trade infringement. So when they work for an enterprise, they end up delivering something that the GenAI model would create that might infringe somebody else's copyright. So they wanted to have this peace of mind that they can use the technology and be protected in terms of potentially infringing. And when we talk to the enterprises, they had the similar risk, but on the opposite side. They wanted to make sure that whatever they're using or whatever they're in production workflows is super safe. And it gives them the peace of mind to use in production and to use in content generation at scale. When we canvassed the world and looked at what was available out there, this was three years ago, there were some open source models, some commercial offerings, they were either not necessarily clear on how they were trained, they were easily generating content that we considered potentially adding those risks. And they didn't have the quality and control our customers expected. And our customers are the most pixel obsessed, the most color space obsessed people on the planet Earth because they really want to tell amazing stories and they need to be in control of that. So after we realized there is no good solution for them, and because we were investing in Adobe research for decades, innovating a lot of algorithms, and as I said, smaller generative models, we've decided it's our opportunity to come up with our differentiated set of models that will cater to the needs of our existing customers, artists, but also enterprises, giving them the peace of mind that this technology is controllable, is safe for commercial use, and it's also developed with the transparency they were requiring. And this is what led to us building our own set of foundational models and doing it in a way we consider makes them commercially safe. Like training on data we have rights to train on from Adobe stock library, designing them to be easily integrated in our products, designing them to be also completely transparent. And maybe we've talked about content authenticity before and content credentials, but we've also innovated for the last five years into actually tagging the content and making sure there is a watermark that highlights how a piece of content was created to have this full transparency in the creative process.

‍

Patrick Moorhead: Yeah, it really makes sense. And I remember the debates on, hey, we may or may not have used YouTube as a training set. We may or may not have used copyrighted images and remove the watermarks. We may or may not have gone on Disney's website, pulled a bunch of characters down to create stuff. And, you know, it's one thing of a consumer. It's another thing if you're a business. I mean, I run a very small business, and I've received, you know, we accidentally used some images and, and for every one of these images, they wanted a $10,000 check. So I can imagine if you're, you know, most companies are a billion times larger than mine, but you can imagine the, the liability this creates.

Alexandru Costin: It. It is. It's a risk. It's a real risk. And, and again, the peace of mind, especially for production, when you actually want to generate content that you publish into the world, it's extremely important. We see even in enterprises for ideation and internal usage, there's more flexibility. We'll talk about that, too, about what type of constraints you want to put on the creative process. But for production use cases, definitely a key factor to consider.

Daniel Newman: So, Pat, you got that letter from me. Did you send the check?

Patrick Moorhead: Definitely was not from you.

‍

Daniel Newman: It was like the prince from a country sending out emails, trying to fish for that. That's like another version of, like, we're just gonna send everyone and somebody's gonna send me the money. I swear, I get those from time to time, too. And I'm like, okay. But actually, I mean, creative and protective rights are a huge inflection throughout this entire. Right, because we know that most of the large language and frontier models were trained on a ton of data that was probably the corpus of data, a lot of it was supposed to be protected. And then fair rights sort of became the. The overwhelming, you know, agreement among the judicial. And now, you know, because Adobe was early and often one of those companies that really stood up and said, we want to be very conscious of protecting the work that's been done by a lot of our designers. Right? And essentially, in some ways, you got a lot of benefit. I remember I said very early, in some ways, it was almost punitive to do that. To take it that seriously. And I'm not saying you were wrong. I'm saying because the market just kind of looked past it, like we thought like the market. But you see now that, you know, I think in the long run, you have to think about enterprise, you have to think about business, you have to think about protecting rights the same way enterprises are thinking about ERP data and about CRM data. But early on it's a bit, a bit of the wild, wild west. Another thing though is like the trend towards mixtures, mixture of experts. Alexandru, so you know your models are great, but there's a lot of companies that have risen, unicorns, competitors, partners and they've become model creators that do something for video or they do something for imagery or they do something for text. I see that with Firefly. Instead of sort of closing the door and saying we're Adobe, nobody gets to come in, you're saying, hey, we're going to just make this all available here. You do it all here and it seems like you're confident they're going to kind of land back on you. Talk about that kind of.

‍

Alexandru Costin: Yes, happy to talk about that. And actually at Adobe, in a previous job gen AI is my latest. I was also in charge of the extensibility platform for the Adobe Creative Cloud products. And we've always had this perspective at Adobe that we need to be an open ecosystem and open platform both in the case of PDF and other standards. We've not only Flash, et cetera, we've always created file format standards that we open sourced and we enabled others to participate in Photoshop in Premiere Pro Extensions, which is this extensibility strategy. We have enabled many companies, startups, or maybe even large companies to add value to our products and build businesses and solve customer problems. By the way, I got acquired by Adobe. I had a small startup in Romania doing extensions for Macromedia Dreamweaver and this is how I got connected to this world too. So this perspective of enabling the Adobe customers at the time, Micromedia customers to tap into additional value by integrating startups' innovation into our flagship products was always top of mind for us and it always is actually listening to our customers and realizing we cannot do and or invent everything that our customers needs. So a great opportunity to bring that choice and bring that richness of options to our customers is by opening up our ecosystem and the example you're pointing out to on the model side. So we did that with extensions, which is like code running in our products, etc. Services integrated with Gen AI, we're approaching it very similarly. So we've launched our models, we're very proud of our models. They are niche for commercially safe generation, controllable, et cetera. But in ideation there are customers expecting the models to do things that are not safe for production, but they're Usable in ideation, in internal settings. So they talked to us and they asked us to start talking to some of these partners like Google, like OpenAI, like many of the startups, emerging startups like Luma or Runway or Flux Labs, and start bringing those APIs into our product line. So customers have the choice of picking the model that has the right personality, has the right understanding of the world for mostly ideation use cases and has the full traceability, the full lineage of edit so they can make decisions. What do I take for production, how is content created and when do I switch to the Adobe models for production? And this is the latest wave of integrations we've been doing since February where not only again continuing revving our models, improving them, I'm going to talk about that maybe towards the end of this session, but also bringing in innovation from either big labs or small labs that are coming up with models with different personalities or characteristics. So our customers always have the choice to create with the best models in the world at their fingertips, either for ideation or production.

‍

Patrick Moorhead: So how does this work? I mean, so you have your models, you have partner models. Is it a, you know, hey, this is safe. These are our models. Partner models are maybe a tbd. How does your customers determine if something is commercially safe? He talked a little bit about it, but how do they know?

‍

Alexandru Costin: Simply, I think right now the mechanism we use is something called content credentials. Let me refresh maybe your viewers memory on what those are for both content editing or generative AI at Adobe we realize that there will be the world. I will do the three seconds. Let me answer again. How do we separate and give full transparency on what models are used for what workflow? We use something called content credentials. In 2019 because we were innovating in AI, we presented at our annual conference a voice cloning technology that was so powerful that actually it triggered a big reaction from the public, a big reaction from the government. And we realized that this technology was very powerful and could enable a lot of harm to be created. So we didn't ship that technology, but what we did in 2019, we realized that we need a world consortium where we can actually collectively start tracking how content was created and edited to give this transparency to the Internet users and enable them to make decisions, if this is a real film or this is a parody, and make decisions on what to believe or not on the Internet. And this has led to the creation of what's called the Content Authenticity Initiative. This is a large consortium with thousands of companies participating. Both gen companies camera manufacturers, social networks, search engines, like a lot, basically everybody that is participating in the Internet digital content ecosystem has joined this initiative and have all agreed to tag and display in their websites content with these content credentials. A digital nutrition label, if you might say, describes how a piece of content was created, edited and gives transparency to the consumer of the Internet content. And this is what we're using today in our tools in the Firefly web applications where you can use Adobe models, you can use third party models, partner models, and they're all tagged with these content credentials. So by browsing over the image, you can see a small content credential icon. When you click it, it will tell you which model was used to create so you can make decisions, safe decisions about which content to use. And this is this traceability and lineage. It's also being discussed to become a law. There's also the EU AI Act. So there's a lot of discussion about making this transparency of how content is created, legislated, which, which we're very supportive of because we've been pioneering it since basically 2019.

‍

Daniel Newman: So we talked a lot about the kind of humanity part of Firefly, and not just Firefly, but kind of the creative realm. In all of Pat's jokings aside, we are all creative in our own way. And you know, one of the things I think that's incredibly cool about what Adobe has done is it's always been focusing on, you know, focused on democratizing creativity, you know, even as it's tools. Because there were a lot of people that became incredible graphic designers on computers that couldn't draw. And by the way, there was a time if you couldn't draw, you couldn't be creative, you know, so this kind of flow and we're seeing this happen, this is kind of one of the great things about AI. But you know, I'm just kind of curious, you know, in your mind, you know, how are you thinking about the future of all of this developing? Because we've got these emerging paths. You got technology, you got humanity, you've got legal and policy, and you're right at this intersection of all of it. But of course, Adobe wants to emerge from the era of the past and what it's built and become a powerful leader for the next four decades. So how are you driving that? How do you see Adobe driving that? And of course and especially in this creative part of your business.

‍

Alexandru Costin: Great question. And we are, this is a key area of focus for us because when we think about the customers we're serving, we have this large base of creative professionals that have spent and they went to school to learn how to master the existing computer vision processing tools with some level of AI. So this is the creative professionals audience which we cater for. And then we have what we call the creator's audience. Creators are the next generation of creative professionals that actually not necessarily work for somebody else as a creative professionals, but they decide many of them to become professional content creators and monetize through social monetize on YouTube, et cetera. So we see this category of emergent creatives that we call creators that will not only create content for the sake of the content, they make a business out of it. So it's a very exciting development. It's kind of what the next generation brings. And we also have enterprises and small businesses that are also using AI to innovate and enhance how they communicate to their audiences, hyper personalizing messages to their users and doing content generation at scale. When we think of all of these three audiences that are using today, various technologies from Adobe and from others, we think of our opportunities to use Genai to help each of them reinvent and thrive in this new world. Let me talk about these three audiences one by one for creative professionals. A lot of the work they do today is kind of manipulating the pixels, doing the operations themselves. And we see the huge opportunity with the Firefly models, with partner models with agentic capabilities to help them move up in the chain. And instead of becoming the visual producers, becoming the creative directors of the content they want to create. Because that's what this technology enables. It's such a powerful human computer interaction improvement that it literally raises the opportunity of how these creative professionals can actually interact with the computer and achieve the goals they have by actually elevating themselves into mini creative directors, if you might say. So this is a big area of focus for us. We're investing in again both elevating the quality and the deep world understanding of our models, bringing in the third party models, but more importantly also investing in this layer of agentic understanding of the user intent and then orchestrating existing tools and AI tools to execute the creative intent. Very excited about Zipros or creative professionals on the creator side. We think they have needs being small business owners. Actually content creation for them is how they achieve their livelihood. But it's also very important for us to help them not only create content faster but do it in a very efficient way, delegate even more to the computer, because they want to also create that scale and do intelligent product placements in their content, because they also need to monetize by doing different brand partnerships. So a lot of effort goes into lowering the bar of entry, enabling this technology to be more accessible, but also expanding out into helping those creators be successful with a larger set of technologies that surround the content creation process itself. Like, we have a lot of marketing insights from our experience cloud business unit, and we're thinking, how do we help these creators actually be successful with the content they create? And of course, for enterprises, this is a huge area of opportunity. The reality is content now is easier to be created than any time before. And this is what happened in previous revolutions like the digital publishing revolution, the digital photography revolution, which each of those big shifts in how easy content creation became. What happened is that jobs didn't necessarily disappear, but the content that was created multiplied a thousand fold. And this is what we expect to happen in enterprises. With this gen AI technology. Being able to create infinite content variations for all the combinations of popular products to target audiences is becoming easier than ever. And it's actually becoming easier than ever. If you can trust the model generated, the content to be commercially safe. Because the world we're going into will be. There will be so much content, you won't be able to police it or to review it manually. With humans, you will have to have the models actually generated on brand. You'll have to have the models generate safe content. And this is how we think enterprises will benefit from some of the innovations we're bringing with the Firefly models, because they are commercially safe, so you can generate with peace of mind. So these are three audiences. And for each of those audiences, the technologies we're building are going to enable either new human computer interaction paradigms to accelerate the exploration process and the production process, or the content supply velocity. And of course, in all of this, our key models play a key role. We also see in ideation, more of the opportunities emerge when we integrate the partner models. Like if you want to create a mood board in Firefly Boards, for example, which is our infinite canvas for teams to collaborate on new ideas, you can very easily go in, select various models from us and from partners, and start ideating and coming up with new ideas so that you know the themes of a particular campaign you want to create later.

‍

Daniel Newman: Wow, there's so much to unpack there and we don't have enough time on the show, but I do want to make a bit of a comment. You really did start to trigger some thoughts about how, you know, Adobe's kind of always had these three businesses that, you know, with the document, the creative and of course the experience that have been somewhat, you know, in their own silos, you could call it. You really do start to see if you're a business, how this could flow, the agents could really flow between these businesses and it would create a much more integrated, you know, continuous experience across the platforms. I think like you create something the way it gets distributed, the way the, the outbound messaging gets tracked, the way an ad gets placed, the way.

‍

Alexandru Costin: It's a huge opportunity.

‍

Daniel Newman: It's a huge opportunity.

‍

Alexandru Costin: We have a group at Adobe, it's the Gen Studio Group. I'm actually meeting them after this interview. We work really closely to really understand how do we bridge the divide between our business units and how do we create this flywheel for our enterprise customers to both not only use the data about their customers they hold with us, we're trusted because we do a really good job protecting their data. We don't train on their data, but we can give them new capabilities to understand their data and automatically use it in the production process. This is the. I was actually part of the team that acquired Omniture. If you guys remember Omniture, this was a startup, a large, medium sized company we acquired back in 2010. And the premise, the thesis of the acquisition was like, let's bring data plus art together and enable this enterprise oriented flywheel where you can actually use the insights you capture about your customers to create personalized content. And I think the technology at the time wasn't there yet. But with this gen AI technologies, with the investment we're making in our models, with the commercial safety guarantee, bringing the enterprise data into the creation process is the flywheel that will help many brands improve how they communicate with their customers.

‍

Daniel Newman: Good.

‍

Alexandru Costin: We're doing many things with Document Cloud too. We call it the business professionals audience. Business professionals using the Acrobat AI Assistant, the Acrobat project. We have many, many new announcements actually happening in that space. But we're enabling content consumption to also be merged with content creation with Adobe Express that is powered by some of these generative technologies that we work on with the Firefly model. So there is a continuum and AgentIQ is a bridge because AgentIQ speaks in English with all of these tools that we have in the Adobe MCP directory. If you might say, I'm gonna go too technical, but there's this huge opportunity to take all the value we have today and expose it as tools to an agentic layer that spans the business unit to create immense value for the users of the Adobe tools across all the business units and audience.

‍

Daniel Newman: Well, Alexandru, I want to thank you so much. Clearly what I started thinking is something that you've been thinking and I'm hoping it's something that all of you out there catch a little interest in and start to look at how all of this is going to work together. Because with these MCP A2A, all these technologies are going to certainly break silos that exist within companies and then again across companies and those that integrate and build the most capable are going to win. But I love being creative. I don't have the skills to draw, but I love that I can think, talk, speak, draw, create. And so often Adobe is the tool that we use to do that. So appreciate you joining us. Let's do it again sometime soon.

Alexandru Costin: My pleasure. Nice talking to you both.

‍

Daniel Newman: And thank you all so much for tuning in to this six five on the road virtual webcast. Great conversation here about creativity. Talk to Adobe Firefly, but we do these all the time so we hope you'll subscribe to be part of our community. Check out all of the content here on list 65, but we got to go for now. See you all later.

‍