Ep 380: Adobe Max AI Announcements – What’s new and will it matter?

Resources:

Join the discussion: Ask Jordan questions on Adobe AI


Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup

Connect with Jordan Wilson: LinkedIn Profile

Try Our Free AI Prompting Course: Register for our free Prime, Prompt, and Polish AI Course! 

Transforming the Entertainment Industry

This innovation isn’t just for commercial video production; it promises implications for the entertainment industry. Analysts predict a rise in the use of AI for Television production, with an estimated 10-30% of b-roll potentially being generated through AI. That’s a significant shift in the technical aspects of non-live entertainment production.

Adobe Sets Precedent with AI Training Approach

Unlike many tech firms, Adobe has based its AI models on in-house data, a measure designed for ensuring privacy, safety, and legal compliance. This approach contrasts with many companies that lean towards internet scraping for their data needs. While some queries have been raised about the origins of a reported 5% of Adobe's images used for AI training, it's clear that Adobe's policy on data sourcing could potentially set a standard for responsible AI development and use.

Major Companies to Follow Suit?

Interestingly, while some major companies and creative agencies have expressed reservations about incorporating generative AI video on a large scale due to data usage concerns, many may follow Adobe’s path. In an industry where responsible data sourcing practices are becoming increasingly significant, Adobe's transparent, in-house focused data sourcing policy may act as an influencer for broader adoption of AI technologies.

Adobe's AI Integration

The announcement comes with exciting partnerships. Adobe is teaming up with Runway, enabling AI video tools to be integrated into Adobe Suites. This integration will allow creatives to import their Runway projects into Adobe’s ecosystem seamlessly. Furthermore, new features such as Firefly video mode have made it possible to generate videos from text and image prompts, offering increased control over aspects like angles and zoom.

Mastering the Craft with AI

The world of AI image generation has evolved significantly over the past three years. While there was initial resistance from many creatives who viewed AI as a threat to human creativity, AI is now an integral part of creative processes. This trend is observed in several creative fields such as marketing, advertising, and video production and is being increasingly incorporated into mainstream media.

Widespread Integration of AI

In today’s digital era, many creatives are using AI unknowingly through popular tools that have seamlessly integrated AI into their systems. The recent AI innovations announced by Adobe, such as improved AI image models and new video creation capabilities, demonstrate how the integration of AI into creative tools can revolutionize the creation of future media content.

Embrace the AI Advantage

The enhancements in AI technologies showcase how AI can be a game-changer for creative endeavors. With Adobe introducing new AI features that are set to bring AI video and photo editing more into the mainstream, the potential for AI to revolutionize entire industries is more prominent than ever.

In view of recent market trends and advancements, business owners are encouraged to embrace the power of AI not only for efficiency but for its ability to transform standard processes and to bring creative projects to life with exceptional quality and versatility.

A Look at the Adobe MAX Highlights

Adobe's conference was brimming with innovation, including the introduction of the Firefly video model, a new generative AI model for video creation. This exciting progress continues to demonstrate Adobe's commitment to pioneering advanced AI technology.

In summary, as AI continues to play an increasingly integral role in the worlds of creativity and technology, Adobe is leading the charge in pioneering responsible AI development. It's a testament to how a blend of AI and creative endeavor can create new possibilities, revolutionizing the ways in which content is created and perceived. These advancements highlight the limitless potential of AI, offering a glimpse into the future of the creative industry. It's an exciting time to be a part of this AI evolution.

Topics Covered in This Episode

1. Adobe AI Announcement Recap
2. Adobe Firefly Video Model
3. Photoshop AI Updates
4. Firefly Model vs Sora
5. Future of AI and Creatives

Podcast Transcript

Jordan Wilson [00:00:17]:
Adobe just announced a slew of new AI updates at their Adobe MAX conference. And here's the reality. For whatever reason, I think when we get big AI updates from Microsoft or Canva or Google or OpenAI, a lot more people talk about these updates, but maybe not as many people use them. Here's what I think is happening with Adobe. This is actually going to bring AI video and AI photos into the mainstream, maybe even more so than all of those other tools and companies that I talked about. So I actually think there's a lot that Adobe just announced at this Adobe MAX that's going to impact us all. So what's going on y'all? My name is Jordan Wilson, and welcome to Everyday AI. We're gonna be talking about everything that's happening with Adobe MAX and their AI announcements and if it's actually going to matter, and I think it will.

Jordan Wilson [00:01:19]:
So, thank you for tuning in. If you're new here, Everyday AI, it's for you. We're a daily livestream podcast and free daily newsletter, helping everyday people keep up with what's going on in the world of AI and how you can use all of this knowledge, all these new tools and techniques to grow your company and to grow your career. So that sounds like something that you'd like to do while you're in the right place. So if you haven't already, please make sure to go to your everyday AI.com. Sign up for the free daily newsletter, where every single day, we not just tell you what's going on in the world of AI news and fresh finds from across the Internet, but we break down our livestream podcast show every single day in the newsletter. So, you know, maybe you're out there walking your dog right now or running on the treadmill. Hopefully, not running your dog on the treadmill.

Jordan Wilson [00:02:08]:
I don't know. Maybe people do that. But, if you miss any tidbits from today's show, don't worry. We always recap it in the newsletter. Alright. Before we dive into today's topic of Adobe, the Adobe Max AI announcements, everything that they, unveiled at their big conference. Let's first start as we do every single day by quickly going over the AI news. So first, Amazon and Databricks have partnered to cut AI cost for businesses.

Jordan Wilson [00:02:36]:
So this is literally brand spanking new announcement, couple of minutes old. So Amazon and Databricks have just announced a 5 year partnership that could significantly reduce cost for companies looking to develop, for companies looking to develop their own AI capabilities. So Databricks will use Amazon's Trainium AI chips to help other companies customize AI models or to create their own, which could lower operational costs. So Amazon claims that its custom chips are approximately 40% cheaper to use than NVIDIA's GPUs, which currently dominate the AI chip market. So this partnership aligns with a broader trend where enterprise technology companies, including Microsoft and Salesforce, are vying for a share of corporate AI investments and or just creating their own chips. Speaking of chips, yeah, we always talk about chips a lot on the Everyday AI Show. And speaking of, yesterday, I was crushing some some chips from my hometown, FYI. But, in the other chips, the GPU chips that power generative AI, NVIDIA is the leader, and their stock has just reached an all time high as major tech companies prepare to report their Q3 earnings and announce their investment in AI.

Jordan Wilson [00:03:53]:
Right? That's what always happens on these quarterly calls. Everyone just says as many AI buzzwords as possible and hope their stock goes through the roof like NVIDIA. So, NVIDIA's shares closed at a 138 dollars a share. Obviously, they went through that stock split a couple of months ago, but that is a 2.4% increase, and past their previous high of a $135, per share back from June. So the company's stock has skyrocketed nearly a 180% this year and has increased more than ninefold since the start of 2023. Weird. The start of 2023 is the one I told everyone that NVIDIA was the most important company in the world, and no one believed me. But analysts attributed NVIDIA's success to its dominant position in the AI chip market where it controls approximately 95% of the market for AI training and inference.

Jordan Wilson [00:04:45]:
Yeah. That's wild. But major tech players are trying to keep up like Microsoft, Meta, Google, and Amazon as they are heavily investing into their own GPUs or other companies as well. So, right now, the, NVIDIA anticipates significant revenue from its next generation AI GPU called Blackwell, which is seeing, quote unquote, an insane demand and has apparently sold out for the next 12 months. Alright. Our last piece of AI news. Google has just announced a partnership, with a nuclear energy firm to power its AI data centers. So Google has recently made headlines by signing a deal with Keros Power to utilize a small nuclear reactor for its AI data centers.

Jordan Wilson [00:05:33]:
Yeah. Microsoft, just announced something similar about last month. So the first reactor is expected to be operational before 2030, marking a significant step toward more sustainable energy solutions in the AI and tech sectors. So this initiative aims to provide Google with reliable power sources while also demonstrating the company's commitment to lower carbon energy options. Alright. Yeah. So, a lot of AI news today, but a lot of the big ones there were about AI, and were about chips and power. And why is that? Well, all of these new tools that we're talking about all the time like Adobe MAX and all of their AI tools.

Jordan Wilson [00:06:15]:
Let's jump into it. But, hey, the reality is these things need power. They need GPUs. Right? So that's why, a lot of what we talk about on the day to day is is really focused on what these big companies are doing to handle all of this generative AI. Alright. So, thank you to our, everyone livestream joining us. Yeah. I know there's been some problems on the LinkedIn.

Jordan Wilson [00:06:36]:
Sorry y'all, but thanks everyone for joining us. Jackie, Brian, and Michael, Melissa, sabbatical, everyone. Tons of people in the house today. So, let's go ahead and get started and give you a quick recap of what was announced by Adobe at their Max conference. So there was a lot, but we're just gonna be focusing mainly, on the, I'd say the bigger picture AI features. Right? Because here's the reality, and and we'll give you a more in-depth rundown, in today's newsletter, but they just sprinkled some AI on every single product. Right? And even personally. Right? Let me tell you my my my personal story.

Jordan Wilson [00:07:14]:
I've been using, Adobe Creative Suites longer than I've been using any other software. Right? I believe I've been using, you know, started using Photoshop back in, 20 2005, something like that. So almost 20 years. Right? So, obviously, Adobe is not new in this AI space. You know, they've been using, different machine learning algorithms, you know, over the last few decades to improve all of their softwares. But now with generative AI, which is very different than, traditional machine learning and artificial intelligence. Right? They're able to offer so much more and to make it so much easier, on consumers who want to take advantage of this technology. So let's just do a quick recap, and I wanna highlight the big stuff.

Jordan Wilson [00:07:59]:
So I'd say the number one thing was the Firefly video model. Alright. We're gonna break that down a little bit more here in a minute. But this is a new generative AI model for video creation, and you can use or you can start with text or images. Right? So, it's right now, available in a limited beta. So there is a, a wait list, at least as of late last night. So right now, it's available in the FireFly web app and in Premiere Pro, which is Adobe's AI video editing software. Alright.

Jordan Wilson [00:08:36]:
So you can get it in multiple places right now, and you do obviously have to have a paid Adobe subscription to take advantage of all of this. So right now, there is no additional cost to use Adobe Firefly, which is great for now, but I wouldn't hold your breath. I watched an interview last night with Adobe CEO who, kind of alluded to that eventually there may be an additional, surcharge to take advantage of this AI video. So we're gonna be doing a kind of, at least, on paper, a comparison of this and Sora. But this is essentially what I think the big news here is. Adobe, is, I would say, bringing AI video to the world's top creatives first. Right? We heard about Sora now 8 months ago, and it is still not yet publicly available. You know, I think other companies have been making great strides in this space, particularly Runway and Pico Labs.

Jordan Wilson [00:09:36]:
Also, Adobe is partnering with Runway, which I think is a smart move. So not necessarily trying to compete head to head with others in the space, but actually allowing creatives to bring, you know, as an example, their runway projects into their Adobe Suites creative, workflow, which I think is a very smart move. Alright. So like I said, text to video. So I'd say that this is the headlining AI advancement is the availability of the new firefly video mode. So to be able to generate videos from text prompts with camera controls, angle motion, zoom, all of those things, and then image to video as well. So instead of starting with a text prompt, the ability to start with an image prompt. So, yes, if you do follow the, you know, AI creative space, this won't seem new on the surface.

Jordan Wilson [00:10:31]:
Right? Because you're like, oh, okay. Well, we have a lot of those features in Runway. We have a lot of those features in Picolabs, you know, Clang. There's all these great AI video tools, and I think that we're gonna start to see, some video capabilities soon here from Midjourney, who is the leader in the kind of, AI photo space. But, yeah, this is all now going to the hands of the world's top creators, and I think that is a big differentiator. Alright. A couple of things. Let's break this down a little bit, in where we're going to be seeing these new features and ultimately what they're going to mean.

Jordan Wilson [00:11:14]:
So this one, I think, is big. The integration in Premiere Pro. Alright. So I'm gonna show a video, example of this a little bit later from Adobe, but I think this is where ultimately, y'all, we're gonna start just seeing little clips and snippets of AI video without knowing. And I think it is because of this one feature, and I think this one feature is probably going to be, the most consequential of anything that Adobe announced. Yeah. They're like, okay. Here's all these AI features in Photoshop and all these AI features in, you know, Illustrator.

Jordan Wilson [00:11:53]:
Right? Like I said, they're sprinkling AI everywhere. But I think it's actually the integration into Premiere Pro and bringing in the Firefly video model into Premiere Pro, specifically for the generative extend feature. Right? I put something out, on on Twitter yesterday, you know, minutes after this was announced, and I said 99.9% of people aren't going to be using this feature, right, this generative extend, because how many of you out there use Premiere Pro? Right? Unless you're a video editor, you probably don't use Premiere Pro. Alright? So very, very few people are actually going to be using this new generative extend feature that is powered by Adobe Firefly. But I will say this, a 100% of us will be experiencing this. Alright? Here's here's the reality. Right? Let's let's just call this spade a spade here, y'all. Everything right now with with Runway and Piccoloabs, right, this is all very experimental.

Jordan Wilson [00:13:03]:
Right? For this to be inside of Adobe's, Adobe's suite of tools is huge. This new, Adobe Firefly video, this is huge because I think this, number 1, it legitimizes the AI video technology. Right? And now obviously having this, you know, even this runway integration built into, Adobe workflows is huge. So I do think, you know, even the top creatives, in the world and, you know, I'm I'm lucky enough to be connected with a lot of, people in the industry. But I think for the most part, there was almost this segmentation, right, where people were looking at, oh, AI video, this is more of an a tool, a toy. You know, we're just using this to explore. We're using it, to ideate, maybe to brainstorm on the front end for storyboards. Right? A lot of companies are using this just to create better storyboards, to pitch projects better, but then, ultimately, they're not always using this AI technology in production.

Jordan Wilson [00:14:01]:
But I think this generative extend feature is going to change that because what it does, like and like I said, we're gonna for our livestream audience, we're gonna show you an example of this. But, you know, let's say you're editing video on a timeline. If there's a big gap or you're like, oh, man, this, you know, this this clip stopped too soon, and we weren't able to show an important piece. Right? Like, our our team stopped shooting too soon or now that we're editing. Right? We see that there's a huge missing visual, and there's a good chance that that video or footage that was shot, you can't just go out and recapture something. So this is, this generative extend feature. Right now, it only adds, I believe, up to 2 seconds of video, which doesn't seem like a lot. But when you're talking about high end video editing, 2 seconds is an eternity.

Jordan Wilson [00:14:52]:
So this feature, I don't think can be overlooked. Alright. And like I said, we're going to be, diving into this, and we're gonna be talking about yeah. Michael's asking here where they get their training data. Yeah. Don't worry. We're gonna be talking about that one soon. And and, hey, everyone.

Jordan Wilson [00:15:07]:
For our our live stream audience, let me know. Do you use any of the Adobe products? You know, we don't actually cover them a whole lot because I think they are a little more niche, right, than using Microsoft Copilot or Google Gemini or ChatGPT. But I think this is going to impact us. Right? So, Photoshop, ton of AI updates. So there is a new generative workspace powered by Fireflies' image 3 model for faster image generation. So Adobe said this is 4 times faster. So essentially, there is a new generative workspace feature, that essentially just brings in this feature into Photoshop. So it's more of a of a user, interface, addition there because you could always access, the, Firefly image model, but this is just more of a seamless way to work it into Photoshop, which, again, I think is going to expose a lot more people to this technology.

Jordan Wilson [00:16:05]:
Right? I still use Photoshop once in a while, not a lot. You know, if I'm being honest, I I use Canva so much more. Right? I used to use Photoshop daily 10 to 15 years ago. I don't really use it a lot. You know, I might use it a couple times a month now, but, you know, it wasn't really bringing their AI image generation to the to the front there. So now this new generative workspace, essentially brings this AI image capability via Firefly 3 into Photoshop. So I do think that a lot of these features that Adobe just announced, you know, it's it's really just making it easier for creatives to use all of this, AI technology where, you know, aside from the video model, there's nothing, you know, shockingly new. But I do like, some of these updates in Photoshop.

Jordan Wilson [00:16:56]:
So it has a new, distraction removal, which I think is pretty big. So, you know, the example that Adobe gave, they showed a photo of all of these wires. Right? It's a building, but there's, you know, a few dozen wires. So, you know, you can one click and remove wires from your photos. So it's it's not new technology. Right? That technology has always been there. It's just been a little more difficult, to use. And, you know, hey.

Jordan Wilson [00:17:21]:
At at least in terms of this generative fill. Right? So what that means is filling certain areas of a photo, with by using generative AI. Adobe's actually been ahead of everyone else in this space. Right? All the other, you know, AI image generators, they just started getting this in the past year. Adobe's been there for a while, with their generative fill features or generative extend. So a lot of those happening in this new update, which I think is is actually really good. And then also, the ability to remove people as well from a photo. Right? We're seeing that now even on our devices.

Jordan Wilson [00:18:00]:
Right? So we're seeing that with Google Photos, Apple Intelligence, when it actually if it ever gets intelligent enough to actually come to our devices and work. Yeah, Apple's been very slow. So a lot of these features are now coming to our smartphones. Right? I'd say Google was the leader in bringing that technology to our our smartphones and then Apple is just following suit, right, where essentially you take a photo. I don't know if someone's photo bombing you in the background or, you know, some, you know, dude is spilling ice cream in the background and it's, you you know, extremely distracting. So this has always been a feature in smartphones. So, Adobe just bringing this. Again, Just updating it, making it easier, one clickability, you know, remove wires, remove people, etcetera.

Jordan Wilson [00:18:46]:
So just making it a little easier to use their generative fill technology. Alright. But here is the big thing that I wanted to, focus on for the rest of this show. There has been so much hoopla and hype on Sora. So if you don't know, OpenAI previewed Sorra a few months ago. Right. And, essentially, this is their AI video model. Right? So they first released this, the preview back in February.

Jordan Wilson [00:19:24]:
So, we've kind of have, quote, unquote, seen this now for 8 months, but it has not been publicly released. I said this when it came out, when Sora was, you know, kind of previewed in a very few, a very limited number of creators actually have access to Sora. I'm I'm sure it's a few dozen, but OpenAI hasn't been very transparent with that. But when this was announced in February, I said, okay. The quality here is scary. Right? It is actually too good to use. And I said that at the time. I said, I don't expect SOAR to become publicly available until after the US election, you know, and now we are a couple of weeks away from the US election.

Jordan Wilson [00:20:12]:
So I would assume that SOAR will be, become available either later this quarter, or early in 2025. And, you know, we shared in our newsletter a couple of weeks ago, they're already working on, you know, a big round of updates for Sora even though the general public doesn't have, access to this. Whereas, Adobe Firefly just beat Sora to the punch. Right? It is live. It is available. It is available in beta. That that piece is important, but it's here. Right? From a quality standpoint, again, I've seen very limited, limited examples so far of Adobe's new, video capabilities inside of FireFly, but Sora is more impressive.

Jordan Wilson [00:21:00]:
Right? But what what does that matter if it's not publicly available? So but I will say this. When working with the generative extend feature of videos, it's extremely powerful. Creating videos from scratch right now in Adobe Firefly, do they look good? Absolutely. Right? If you would have told me a year ago, hey, check this video. Is this impressive? If what if I told you it was created by AI? I'd be like, oh, freak yeah. That's really, really good. But it's still not as good as what we've seen from Sora. But, hey.

Jordan Wilson [00:21:34]:
What does it matter if that's not publicly available? But, I think there was actually some big pieces here that weren't talked about. So first, let me just give you a brief update on Fireflies. So, it actually launched last March for image generation, but as of yesterday, now it has video. So there is both you know, if you're hearing Firefly AI from Adobe, it is it was a image model up until today. So it doesn't have a different name. It is just, you know, Firefly AI for video. So an important distinction. So this is the first publicly available video model for commercial use from Adobe.

Jordan Wilson [00:22:18]:
And also like we talked about, you can start in a variety of ways. You can just use a text prompt. You can use a photo prompt or inside Premiere Pro, you can just extend your current video, right? So there's 2 different pieces there. You can create something new, create something new with text or with a photo, or you can extend something that already exists. And if I'm being honest, I think a lot of people are talking about the first part and not the latter part. Where I actually think the latter part, even though it's not incredibly sexy or or cool or, you know, grabbing headlines, that's where I think a 100% of us are going to be exposed to AI, AI video almost immediately with this generative extend. Alright. So right now, like I said, the generative extend is available in Premiere Pro Beta right now, and it allows users to extend video clips by up to 2 seconds while maintaining camera motion and subject movements.

Jordan Wilson [00:23:21]:
Alright. Also, it needs at least 3 seconds of video to work with. That's important as well. Right? You can't just, you know, put a one second video clip in there, and it can create more versus what it doesn't have. So you need to give it enough, information for it to actually extend the clip. Alright. So let's go ahead and quickly look at this feature live. Alright.

Jordan Wilson [00:23:45]:
So, always love doing live demos here on the show because they always go right. That's a joke. They generally always go wrong. Alright. So, hey, livestream audience. Let me know if you can, see the screen here And for our podcast audience, so what I'm going to be showing you here, this is from Adobe. Alright. So what we have here is, essentially, they are showing a video here and, saying, hey.

Jordan Wilson [00:24:13]:
Meet generative, extend, more frames, less frustration. So, what they're trying to portray here is there's a man in a diner, it looks like in a desert, and they're saying hold on a moment. Alright? And then they're zooming into the timeline. And, this is every video editor's worst nightmare. Alright. Small known fact, about me. I I worked at a nonprofit for 10 years. We just became an activation and creative agency for Nike and Jordan brand, but I spend the majority of my time editing, videos for Nike and Jordan, right, on our partnership projects.

Jordan Wilson [00:24:53]:
In this scene right here is every video editor's, every creative's worst nightmare, where essentially you have either an audio track or you have, you know, maybe some dialogue and you have this huge gap in your visuals. Right? Sometimes, I would say a lot of times for me, this is when you are interviewing someone and you have their dialogue going, and you're trying to show different b rolls. So b roll is essentially something that you splice in. It's video that you splice in to make interviews more interesting. Right? But this also comes up outside of just interviews. Right? If you're putting together a feature film and you're like, oh, man, or or a commercial, a 30 second marketing video for your company, there's so many use cases where you're just staring at this, you know, this gap. And you're like, oh, man. We didn't shoot enough or, hey.

Jordan Wilson [00:25:47]:
The story went in a different direction versus what we thought, and now we just don't have a certain visual or a certain scene that that we really need to connect the dots. So what would normally happen here is you might go and use some stock footage, and that stock footage might be terrible. You might just reuse a shot that you've already used and hope the audience doesn't notice. Or sometimes you're just like, alright. Well, we're just going to, leave these dots disconnected. And hopefully, the audience, they're not impacted, and they can't tell that we essentially really could have used a visual here as a segue between 2 different stock shots or, between an establishing shot of a scene and then a close-up of a person. Right? You might wanna have something in between like a mid range shot. So you have this gap.

Jordan Wilson [00:26:34]:
So this is where this new generative extend feature comes in from Adobe. Alright. So now, essentially there is a toggle switch that you can click this new generative extend and it changes your cursor more or less. And then you'll see in this video here, all they do is they have this first shot, where this man was kind of looking off in the distance and there was a gap. There was about a 2 second gap. So now all they did was click on this generative extend. They grabbed the end of that clip and extended it until when their next clip was. So they essentially filled in a 2 second gap.

Jordan Wilson [00:27:10]:
Alright. Then it says preparing. So at this point, most of the videos and reviews that I've seen said it may take up to 2 minutes early on. Alright. I've seen anywhere from about 30 seconds to 2 minutes. I haven't personally had a chance to do this yet. Like I said, I'm not, you know, I'm not making videos every single day. I'll probably have, my my my colleague check this out and maybe do, do a video on it just because he's usually editing more video than I am.

Jordan Wilson [00:27:39]:
Alright. So then once it generates, so now it's showing essentially what is here is AI generated. Right? So, it essentially studies the scene. Right? It it uses advanced AI and kind of gobbles up. Here's what was happening up until this point where it cut, and then all of that is AI generated. Alright? I'm gonna let this I'm gonna keep this playing for a couple of seconds to show one more example because that one, you might not have been too impressed. Right? Essentially, that first clip ended with a man looking into the distance, and that's all it really added. So you could make the argument that it didn't really create or extend anything.

Jordan Wilson [00:28:22]:
It just kind of delayed it or made the duration longer. Right? Another thing video editors always do when there's a gap like that is they might just crank something up in slow motion or play something back but in reverse. Right? There's a lot of tricks that video editors use, but here's another one. Right? So this clip ends, the same man who was in the desert. He's kind of stumbling around a gas station. And now it looks like the clip ended as he was walking toward a car. Alright? That can be hard to show a subject in a video because think think. If it's very cinematic, things like your depth of field might be changing, your focus is changing, the the exposure, the composition is changing.

Jordan Wilson [00:29:05]:
This is a much more difficult shot in theory for AI to extend. Alright. So similarly, the in this example, the normal video stops there with a man approaching a car. So then using the generative extend, it keeps going. Alright? Audience, let me know. Does this look does this look pretty impressive and real, this this generation here? So right after this, the AI generated, he still is takes another couple of steps toward the car. I get what you're saying. What's the big deal? This doesn't seem like a lot.

Jordan Wilson [00:29:42]:
Right? This, I think, is going to bring AI video to the main stage. Alright? This small little feature this small little feature of extending videos by 2 seconds. Right? And and to add frames and to continue in the action. This is actually huge. This is huge. And I think that the majority of of TV shows, essentially, any entertainment that is not live, you're gonna be, quote, unquote, seeing this almost immediately. This is huge for video editors who oftentimes will spend countless hours just trying to fill in those little gaps. Right? Oh, okay.

Jordan Wilson [00:30:36]:
Well, which shot are we gonna reuse? Oh, great. Now we have to go get the crew out and reshoot something. Not anymore. Right? I could see, you know, 10, 20, 30 percent, of certain, you know, b roll now being generated with this new feature. So this is actually pretty big. Alright. Let's get out of the live, and let me talk about what this means. A couple of you were saying this earlier.

Jordan Wilson [00:31:06]:
Yeah. Josh says brings a whole new meaning, to the comments, you know, common phrase you hear. Hey. We'll do that in post. Yeah. Now in post, they'll just create new, clips of scenes. Right? So someone earlier was asking about training data. Right? This was a straight shot.

Jordan Wilson [00:31:28]:
I don't think a lot of people caught this shade that was thrown by Adobe. Let's talk about this. So Adobe made a very big deal, and rightfully so, that they do not scrape the Internet. Right? They brought this up many times, in their presentation yesterday at Adobe MAX. And by the way, the Adobe MAX conference, goes, also today and tomorrow, but I think all of the headlining features, they obviously announced during their keynote. So in case you're wondering like, hey, why is Jordan talking about this one is still going on? Well, I think that's essentially the big pieces, the the the big announcements that we're gonna get. But I think a lot of people skipped this. Adobe very explicitly said multiple times that we do not scrape the Internet.

Jordan Wilson [00:32:20]:
Alright. Really a focus on safety and the legal ability to use this. So, Adobe says that their, their models are trained on in house. Right? It's all of the content, that they have access to. Right? Not scraping the Internet, and that's huge. And I think that's one of the reasons why some of the biggest companies in the world, some of the biggest creative agencies, some of the biggest advertising agencies have not yet rolled out, generative AI video and photo, to the largest clients. I think for the most part, y'all, for everything else, for all the, you know, medium sized clients, these companies have been using AI video, AI photos very frequently. But for the biggest commercials, right, the commercials you see on the big sporting events, the big news events, whatever it is you're watching on TV, I think for the most part, we haven't seen this kind of mainstream, AI video, but I think now we may.

Jordan Wilson [00:33:25]:
Because one of the biggest, you know, elephants in the room is, okay. Well, can we actually use this? Right? In some of the bigger companies they offer, like, as an example, I think Microsoft and Google, they give you kind of some blanket protection. Right? And, you know, companies don't really disclose how their models are trained. So that's a huge benefit that I think Adobe has is they're saying, hey. We don't scrape from the Internet. All of the content that we use in our Firefly model, is from content that we have access to. Alright? Now I do have to call this out, because there was an instance where, it looks like, okay. There was some some midjourney images that made their way into, Adobe's trading data.

Jordan Wilson [00:34:18]:
So, you know, I'm not just gonna, you know, spit back to you, exactly what the company says. No. There's there's been some discrepancies, that have come out. So it's like, okay. Well, if you're using other, AI photos or videos in your training data, does does that mean that you're not scraping the Internet? I don't know. So, yeah, previously, there was a report that about 5% of the images used to train Adobe's Firefly AI tool were generated by other AI platforms, including Midjourney. So I think that piece is important to call out. So, aside from that, though, there hasn't been any other reports or any other news or information, that goes to contradict this statement from Adobe.

Jordan Wilson [00:35:08]:
And I think this is huge because I think you do have to, be take this, the training and the data privacy part, that's huge. Right? Because you don't know in the end. Right? You don't know in the end if you're using MidJourney, if you're using stable diffusion. Right? So many of these models, these AI image generating models, you don't know a lot about their training data. But for the most part, it's just kind of assumed that they're trained on the open Internet, that if something is available out there to humans, that these other companies are essentially scraping everything and training their models off what's already out there and not really paying much attention to copyright. Right? Because, essentially, the argument that these other companies are making, well, hey. If we just scrape every single visual or every single video that exists, hey. When we're using this kind of diffusion, technology, right, we're not actually recreating anything that's copyrighted.

Jordan Wilson [00:36:12]:
We're just, you know, pulling bits and pieces from millions of different pieces of copyrighted work. Right? So Adobe is a little different here, very explicitly saying we do not scrape the Internet and, responsible AI training. And this is, I think, a huge direct shot at OpenAI and Sora even though Adobe didn't straight say this by name and, you know, call Sora out. I mean, that's what was happening here. So there was the the infamous, interview with the former chief technology officer of OpenAI, Mira, Mira Muratai, who in a sit down interview with the Wall Street Journal, and I think this was one of the things that kind of exploded, Soera onto the scene for good and bad reasons. Right? I think this this interview this sit down interview with the Wall Street Journal really brought Sora to the non AI enthusiasts. Right? But there is a big flub here in the interview when Mira was asked a fairly straightforward question about the training data, and she didn't have a good response. She more or less just said, I'm just not gonna go into the details of the data.

Jordan Wilson [00:37:30]:
Right? Softball question. Right? It wasn't a hard hitting, like, hey. FYI. You know, here's a gotcha question. We sat down and analyzed, you know, hours of Sora video and found x percent was you know, it wasn't a curveball. It was a very softball question about, hey. Tell us about the training data. And, OpenAI didn't really have a good response.

Jordan Wilson [00:37:53]:
Almost like she was caught off guard for the most basic question that you could be asked in that situation. Right? You're showcasing the power of Sora video for the Wall Street Journal. You gotta be able to, you know, answer some very basic questions, about the training data. So, pretty interesting shots fired there from Adobe and essentially saying, hey. We not only is our a you know, Firefly AI video model now available unlike OpenAI's Soarer, but we can answer questions about the training data because they say it is all based off of data that they have access to on their platforms and not scraped from the Internet. Alright, y'all. So let's let's wrap this thing up. So what does this all mean? What does this all mean? Well, a couple of things.

Jordan Wilson [00:38:50]:
Number 1, I think there was this initial wave of creatives. Right? Probably 2 or 3 years ago when this generated you you know, the ChatGPT boom and, you know, actually, DALL E, the early versions of DALL E, you know, they weren't any good, but that actually preceded ChatGPT. So, you know, AI image generation is is not new. It's been, you know, pretty popularized and mainstream now for 3 years. It's obviously gotten much, much better, over the last couple of months, but I still think there's been this this facet, this group of creatives that have said, no. We're not gonna use AI because I think originally they thought of it as a threat. Right? They could see these videos that AI could produce, Right? These Sora clips that were, you know, 15 seconds from a text prompt, these great AI images that you can start with in Midjourney and then take that into, you you know, programs like Runway. I mean, you saw literally short little films being entirely created with AI tools like Midjourney, Runway, Piccoloabs, Stable Diffusion, etcetera.

Jordan Wilson [00:39:55]:
And I think you saw creatives. Right? And there are so many other applications I'm not even talking about here that are now possible inside of, the Adobe, Creative Suites prod, prod products, right? So things like like logo, like logo design, like editing photos. Right? Now you can use generative AI in every step of the process. So I think originally there's this group of creatives that said, nope. Never. You know? AI is never gonna be able to do what I can do. You know? You know? I'm never gonna use AI. Right? Human creativity cannot be cannot be matched.

Jordan Wilson [00:40:31]:
Well, yeah, it can. I've been saying this since day 1, y'all. I go I I have the receipts, 400, almost 400 episodes on your everydayai.com. I've never veered away from this that AI is more creative than the most creative human because it is trained on the exist the entire existence of human creativity. So I think early on, it's just people that don't understand AI and, you know, they take one screenshot and they say, oh, look how bad this AI photo is or look how terrible this AI video is. Well, a lot of times, you don't know what you're doing. Right? And a lot of times, people are manipulating and sharing some of the worst examples just to make themselves in their careers feel more secure. Right? Like, oh, look at this terrible AI.

Jordan Wilson [00:41:13]:
Companies can't be using this. Yeah. They can. So if you are a creative and if you're in marketing, if you're in advertising, and you've still been on this anti AI bend, you can't anymore. You can't anymore. Good luck, all you anti AI creatives that said, I'm never gonna use this. Because guess what? What are you using? Are you using Canva driven by AI? Right? Are you using Adobe? Now AI is in every crack and crevice of Adobe. So if if if you are a creative that still thinks, oh, I'm not gonna use AI.

Jordan Wilson [00:41:48]:
AI is not as creative as humans. Well, good luck. Good luck using all of those day to day creative tools that you rely on that are now just powered by AI on the front end, the back end, and everywhere in between. You don't have a choice anymore. You don't have a choice. So good luck. Jokes on you, all you people that said AI is just a joke. It's not actually real.

Jordan Wilson [00:42:12]:
Well, okay. Most most feature films and, right, like, many of them start somewhere in in Adobe product. So you're you're using AI whether you want to or not. And then I think last but not least, this is a huge step between blurring the lines between AI in the real world and bringing AI video mainstream. And I think, like I said, even though it is very small, that generative extend is going to be widely used. And unbeknownst to the majority of humans out there, so many of these short little scenes, short little clips in our favorite TV shows, movies, etcetera, they're gonna be fake, and you're not gonna know. And I think, for the most part, that's okay. Right? I think at least with what we saw from Adobe, this isn't changing the narrative.

Jordan Wilson [00:43:04]:
This is just making the narrative a little better with generative AI. Alright. I know that was a long one y'all. I hope it was helpful. So, Adobe Max AI announcements, a lot going on. We gave you a quick run through of some of the, more, I think, some of the more highlight worthy features, in Photoshop. The ability to quickly, in a little bit more intuitively clean up your photos and all of that. Yes.

Jordan Wilson [00:43:32]:
We have improvements to the AI image, you know, Firefly model, but I think the big one here is the video model. And although, yes, it is now available to, you know, create, short video clips with text or with, photos, I think the big one here, yes, it is number 1, a shot at SoRa, but number 2, this is making AI integration so seamless, that we are going to be probably, I think, within a year or 2. Everything we watch will have a large percentage of AI video that actually never happened. Alright. Thank you for tuning in, y'all. If this was helpful, please share it with someone. Please you know, if you're on LinkedIn, please tag someone. If you're listening on the podcast, please subscribe on Apple, Spotify, wherever you get your podcasts.

Jordan Wilson [00:44:21]:
Follow and leave us a rating. And when you're done with that, if you haven't already, go to your everyday ai.com. We always leave the link in the show notes where we break down today's conversation. I hope this is helpful. Hope to see you back tomorrow and every day for more everyday AI. Thanks y'all.