About the AWS x Crop.photo Webinar
Are you keen to understand how to automate your brand's bulk image editing workflows with Amazon Web Services’ AI? We recently hosted a webinar featuring Rahul Bhargava, CTO at Evolphin and Crop.photo, and Kevin Carlson, Global Head of BD at AWS. They shared practical insights into bulk image editing for eCommerce, sports & other industries, straight from AWS & Crop.photo's revealing case studies.
Ready to save time and money in your bulk image editing workflow? Watch the recorded webinar.
Transcript Available
For those who prefer reading or need to quickly reference key points, we've prepared a full transcript of the webinar. The transcript covers the in-depth discussion and answers to audience questions, providing an invaluable resource for anyone interested in the topic. Check it out:
---
**0:01**
Hello and welcome to the AWS and Crop.photo AI webinar. I am Rahul Bhargava, and I'm the CTO of Evolphin and Crop.photo. Joining me today is Kevin Carlson, the Global Head of Business Development at Amazon Recognition. Before we start, I just want to let you know, if you're having any trouble with audio, please check your Zoom preferences. So, with that, let's get started.
**0:36**
Today, we are finally going to answer the question: How do you unlock hours and hours of creative time using AI? Before I get into the webinar, I just want to let you know, if you have any questions, please feel free to send them ahead of time, and that way, we can queue up the question. We will basically answer them at the end of the webinar.
**1:01**
So, the agenda today is to review the challenges that exist with manual bulk image editing when you are working with a large volume of images, be it 10, 100,000. We review the challenges, we look at the ability of AI to address these challenges in post-production, and we will have a live demo of the various AI technologies that work in conjunction with the Crop.photo offering with AWS Recognition. We will also have a demo of a generative AI use case, and then after that, we will have Kevin talk about AWS Recognition's AI capabilities and how they help our target users. And finally, we'll have a roadmap discussion on what's next for generative AI and, in general, Crop.photo's AI. And finally, conclusions. There will be a question and answer session at the end, and if you stick around till the end of the webinar, there will be some freebies that we'll be giving away.
**2:10**
So, let's talk about the challenges of bulk image editing. The key issues that you probably are aware of, if you are in any sort of post-production scenario, whether you are creating images for a sports website or for retail or apparel, and any number of industries, is the turnaround time. To deliver high-quality edits is demanding; it requires time, and especially when you have tight deadlines and events, it can be extremely stressful. And then, secondly, the cost. So, if your retouching team or your offshore retouchers are spending several minutes per image, this all adds up when you are editing 10, 50, 100 images. And finally, the production capacity. So, you might have a team set up offshore, you might have your in-house team, but if you have a major event or sales, then you have a hard time scaling up your production capacity. You might be all set to produce a couple of hundred images a week, and then all of a sudden, there's a requirement to do 300, 500, or so. That's where the image editing workflows choke.
**3:25**
Now, one of the things that I want to point out is the typical marketing tasks that are involved with preparing images that go on your CMS or publishing or print. You need to crop images, you need to align, resize objects in the image, you often need to replace the backgrounds or extend the backgrounds. These are examples of low to mid-end image operations that creative departments do day in and day out. And if you look at the current way of doing things, you could be using Photoshop and spending tons of hours doing that, or you might be sending your images to an offshore service, and that can get pretty costly as the number of images grows and the time to completion is an issue. So, as we go through the webinar, you will see how Crop.photo generates an immediate return on investment by lowering the amount of time to minutes to edit several hundreds or thousands of images and costs a fraction of the amount.
**4:32**
Now, in terms of bulk image editing, Crop.photo has a lot to offer, and I'll be going through all of this in a live demo very soon. But just to highlight, in e-commerce or sports or advertising and marketing, there are common tasks, and Crop.photo can work with various image types, whether they are flatlays, product images, on-model images, or lifestyle images. It offers AI image upscaling; you can use object and face detection to recognize objects or facial features, to trigger body-aware cropping or headless cropping. You can use AI to recognize the contour of products and automatically add shadows. If you are working with banners and you want to extend the background, to do outpainting on your images, you can use generative AI. So, there are a number of tools that Crop.photo has, and the workflow is quite different from a normal editing tool. So, with a normal editing tool like Photoshop, you open up the image, do the operations, and then you open up the next image, and so on and so forth. With Crop.photo, as you will see shortly, you define your settings that you want to apply in bulk to all these images, then you upload them in a batch, and then you process them all in the cloud.
**5:58**
Before I get into the demo of the product, I think it'll be good for us to take a look at some of the users who are right now using Crop.photo. We have over 5,000 users on our platform since we launched in the middle of last year. And I'm just going to highlight a couple of them. One of them is Hemline. So, this is an SMB that has over 100 plus brands that wholesale product images to and products on their website. So, Hemline had the challenge of a lot of new SKUs coming in; they were handling over 800 new SKUs every month, and that led to 3 to 5,000 images being cropped every single month, with sometimes 20 to 30 jobs happening a day, and their offshore team was not able to scale up. So, they migrated all their image post-production to Crop.photo. Some of the automations that they use include our body part-aware resizer and our auto-align and resize automation. So, they work with both on-model and regular product images, whether it's jewelry or accessories. And what they are seeing is they can easily process several hundred images a day, and what used to take over a day or two can now be done in a couple of minutes. So, overall, benefits to them were their productivity went up significantly, over 70%. They also solved the problem of inconsistent VPI (vendor-produced imagery) that would come from various brands. It would come in all different resolutions and aspect ratios, and they needed to create uniform aspect ratios. So, they use our AI scaling to automatically ensure all images look great. And based on the ROI that they are seeing, they think they're saving over 250 hours every month.
**7:23**
Now, let's take a look at another case study. This is a company called Workwear Group, based out of Australia, and they had this interesting challenge of migrating from one e-commerce platform to another. They were going to a SAP e-commerce platform. So, this company produces dresses for companies. So, if you are a bank or an airline and you have a uniform for your staff, they will create a portal where you can go and order dresses for the staff. And so, for each company that they work with, they have a portal with lots of images to show the dresses and the choices. And so, they had the challenge of migrating 700 client portals with their imagery to their new e-commerce platform, and they were looking at migrating around 70k images. So, using Crop.photo's lifestyle automation and auto-align and resize automation, they were able to achieve that. And I think they reckoned that they are saving seven figures. They had quotes from other companies to basically automate this migration, and they're saving, conservatively, at least 400 hours with Crop.photo automating routine image tasks.
**9:28**
Now, before I go to the demo, I should also point out that Crop.photo is a standalone AI service, but it can also work in conjunction with a DAM, a digital asset management system. So, we are in the process of exposing APIs and SDKs that would allow pretty much any DAM or MAM to offload images directly from, let's say, a folder in the repository and push it into Crop.photo. So, this is an example with our Evolphin Zoom DAM, where a marketing person could select, let's say, all the shoe imagery. You can see they have different backgrounds and non-uniform sizes, and they could just invoke Crop.photo and apply a recipe that they define on Crop.photo, like align shoes at exactly the same level with a uniform background.
**10:20**
All right, it's time for a live demo, so I will just switch my screen. Give me a second. Here we go. Okay, so this is the Crop.photo web app. It's completely cloud-based; there's nothing to install. And, as you can see on the homepage, there are a number of automations that are defined for you that will be set up in your account, and you can click on these automations and start using them. So, there are a lot of automations; I won't go through all of them. I'll start with one of the automations to illustrate how it works. So, we'll take one of our most popular automations, which is a headless crop, that is used in the fashion and apparel industry. So, if I click on the Headless Crop, I would have the ability to then drag and drop the images that I want to work with.
So, let's say these are the five images. I'll just drag and drop the images into the platform. Once these images are uploaded into Crop.photo, then I can go ahead and define the settings that would apply to all of these images in bulk. So, since I'm interested in a headless crop, I'm going to choose a marker for face, let's say, between eyes and nose. So, Crop.photo is using, behind the scenes, AWS Recognition to identify the faces using facial recognition and pulling in the various markers. So, I can choose from any of the body markers or face markers that are in here. So, a common one is between the eyes and nose. I can select the crop area, and then I can choose a background that I want to apply. So, if I don't want a white background, perhaps I want a gradient image background, I can do that. I can set up a transparent background, so it's going to use our segmentation feature with our AI to remove and replace the background. And later, you will see we have even the ability to extend the background if you don't have sufficient pixels. And finally, you choose the various output sizes that you desire. So, let's say I was doing this for Shopify, and I needed a 2:3 aspect ratio. So, I can set it up to go 670 by 1000 pixels, and I can add as many sizes as I want in here. So, that's the beauty of Crop.photo. With a single automation, I can target many different marketplaces, many different channels, social media, etc. I don't have to crop them again and again. So, once I do that, select the output setting, I can review, make sure everything looks good, and then I just hit "Start Crop." And this launches the task on the AWS backend. It'll fire up the various AI models to start processing these images. So, with our service, unlike other editing services, there is a startup time for any batch. There's a minimum startup time of 30-40 seconds, and it basically then scales up pretty nicely. So, you will begin to see the speed benefit once you have over 10 images or even 100 images or a thousand images. It'll scale up quite fast. So, I'll basically speed up the demo to show you the result of this crop. So, once the crop completes, you will see the status in the projects that you have, and it will say completed, and I can click on it and see the result of this headless crop. So, if I click on these images, I can see the before and after. So, this image was in an aspect ratio that was this, and in this case, I have cropped it to 1080 x 1080, which is a square aspect ratio. So, it's automatically recognizing the facial markers, like between the eyes and nose, and it also recognizes the entire body to auto-align the image within the canvas. So, you don't have to go back into Photoshop and then stretch the image or try to do a best fit inside the canvas. So, all these images are now cropped at 1080 by 1080. It can also... You can see the great job it's doing removing very funky backgrounds. I think it's probably got one of the best background removing capabilities in the industry. A lot of folks who have used Photoshop have said it's even better than Photoshop. So, this is a very low contrast image with regards to the white dress and the background, and you can see how well it does that. It's also able to handle when the face is not looking straight at the camera. So, in this case, it's a side angle. If you look at this image, there is basically a back pose. It's still able to detect roughly where the head and where the eyes and nose are and then perform a crop. So, that's basically it for this crop. And once you have finished that, you can select all the images and hit "Download" and bring them down into your desktop and publish them into your CMS or whatever you need to take them. So, similar to the headless crop, we have other crops. I'll talk about lifestyle crop, which is used a lot by a variety of users from manufacturing to retail. So, the idea with lifestyle crop is for AI to recognize the object of interest. So, behind the scenes, it uses various types of AWS Recognition APIs to detect objects, to detect on-model subjects, and it then zooms in on the subject to basically create a close crop of the subject. And the best way to show you this is to actually go to a crop that I have completed. So, let's take a look at this particular crop that I completed, and I'll pick one of the images here. So, in this case, we can see the focus was on the bag, and without any prompts, without any human intervention, our AI recognized that the subject of interest is a bag and zoomed in, converted this aspect ratio from portrait to a 1080 by 1080 square aspect ratio. And you can see the bag is perfectly in focus; it's not cropped out. So, this is a lot smarter cropping than what you would do with manual editing tools, which don't have awareness of object cutouts, etc. This is all automatically handled by our AI. Here's another example. So, in this particular shot, you can see the object of interest is these books and sunglasses. They're aligned to the right, but our AI was able to recognize the objects in here and center them and zoom in with the margins that the user desires.
**17:58**
All right, so that's a lifestyle crop. And let's take a look at banner production. That's a very common task. A lot of folks need to generate banners for social media or for paid campaigns like Google Ad or Facebook Ad. So, let's see how Crop.photo can help with that. So, I have, let's say, a single image, and I want to generate multiple banners from this particular image. So, this image has a limited amount of pixels for some of the banner sizes that I'm going to show you. So, in this case, a user can come in and turn on our generative fill. So, this is going to use the generative AI behind the scenes, and you can add prompts. This is optional. We also have an auto mode where you don't have to set and try out different prompts, but in this case, we added a prompt to extend the sky at the top and extend the sand at the bottom of this particular image. So, to speed this up, I'll show you the final result. So, once this crop was completed, so you can see, we generated several sizes in here, and the particular size that required the background to be extended using our Outpainting APIs with generative fill have been marked with this label "Gen AI." So, I'll pick one of these to show you the before and after. So, this is the actual image, and we can see on the right-hand side, it has extended the sky and the beach on the left and right. So, there weren't sufficient pixels, but since the user wanted a 1200 by 1000 aspect ratio, which is much wider than this image, the Crop.photo tool not only resized the image, zooming in on the model but also did the expansion of the image on the left and right. Let's take a look at another size. So, this is a generative AI with a very tall aspect ratio. So, the user wanted this to go to 480 by 2400 pixels. So, here, behind the scenes, several things are happening. The original image was AI upscaled, first of all, because there were insufficient pixels. It was a low-resolution image. And you can see at the top, a sky has been added, and at the bottom, the beach has been extended, and sand has been added. So, that's the power of our lifestyle automation that allows you to not only smart crop but also extend. Here's another example with the automotive industry. So, suppose you wanted to extend this car for a very wide aspect ratio, as you can see here. Let's say you were doing an ad for Google, where there was a really wide banner. So, on the right-hand side, we can see it has extended the background, added all this stuff here on the right, to create, and even on the left, it has extended the guardrail. So, now, you can get a perfect banner from this. So, besides this, there are other... Headshot is basically another very popular... Let me go to the sports example here. This is a very popular automation using our headshot fix automation. So, a lot of sports teams use us to create perfect crops for game day. One of Europe's largest soccer organizations is using us to essentially take a set of images that a team might send for the various players, and often these images arrive very late in the season, sometimes like an hour before a game is about to start. They will get 30-40 images of all the various players, and they need to replace the background and crop it perfectly to put it out on social media or on their website. And they can use our headshot fix to essentially recognize the various facial dimensions, including the height of the player's face, and then create perfectly aligned crops.
**22:20**
So, that in a nutshell is Crop.photo. There's a lot to show, but I will now stop this part of the demo and proceed now to AWS Team to talk about the AWS Recognition and how it helps Crop.photo-like services. So, Kevin, please go ahead.
**22:43**
Yeah, thanks, Rahul. Really exciting use cases you can see in Crop.photo there. I wanted to spend just a few minutes talking about the technology that's behind some of the really innovative things that Crop.photo is doing. And the first of those is something called Amazon Recognition. So, Amazon Recognition is a fully managed AI service, meaning that it's accessible as an API, and it does quite a few different things. Some of them are listed here. You can see that it can moderate content, so it can look for things that might be objectionable in an image. It can do quite a bit with face detection and analysis, and we'll talk about that in more detail. There is a labels API that can return thousands of objects and actions and information about a scene or an image. It can actually extend Amazon Recognition to find things that are of interest to you. That's a function called custom labels. We can detect text in an image, so, you know, not your traditional OCR, but rather free text that might appear in an image like the example here, which is numbered bibs in a marathon. We recognize thousands and thousands of celebrities natively, and we have some specific functionality that's built around video. But this is a fully managed service, and these are pre-trained models, meaning they're trained on, you know, tens and tens of thousands of images, and the result of that training allows us to deliver these answers as a service, and that's what Crop.photo is using as a baseline to deliver a lot of this advanced functionality. Next slide, please.
**24:13**
So, let's talk about facial analysis and some of the facial landmarks that are being used. Amazon Recognition delivers over ten attributes and about 30 distinct landmarks for a face. And these include things like the appearance of age, the appearance of gender, the appearance of emotion. You know, we say "appearance," of course, because we can't interpret the actual things, but this is what the image appears to show. We deliver these 30 different facial landmarks, so the left and right corners of the eyes, nose, pupils, mouth, etc. These are the facial landmarks that Crop.photo then uses to make decisions, intelligent decisions, about what a user might want to do in a cropping action. We can also look at things like brightness and sharpness, general attributes. Is the face occluded? Are there sunglasses? Are the eyes open? Things that are important in a photography setting to determine whether this is a good photo or maybe a photo that isn't as appealing. And then we can also look at the direction of the face, so pitch, roll, yaw, and then eye direction, pitch, and yaw. And so, all these signals, you know, on their own are maybe not so useful, but when a partner like Crop.photo takes them and combines them into something that's targeted at a specific industry or specific function, you know, it becomes very, very powerful, as you can see.
**25:30**
And then, just a word about generative AI. Right? So, I said that Amazon Recognition is a set of pre-trained models with a desired outcome. Everyone's heard a lot about generative AI and about large models, and so, at Amazon, our approach toward generative AI is primarily through a service called Amazon Bedrock. And this takes the concept of choice in model provider. So, you can see we offer a lot of large models from a number of providers, some our own, called Amazon Titan, but also from Anthropic and Meta and Stability AI, 21 Labs. And these models are all good at different things. What Amazon Bedrock does is offer them as a service so that rather than getting deep with a particular model, you can have sort of a consistent way of accessing these models and having them help you with your business. And the kinds of functions that Rahul showed, like extending an image and generating content that might not be in the core image, Amazon Bedrock is the service that offers you access to tools that can help do that, as well as do quite a bit with text that might appear inside the image or associated with the image. And I think you'll see a lot of more exciting things through Crop.photo using generative AI as time goes on. And I'll be happy to answer questions about any of this at the end of the webinar.
**26:54**
Thank you, Kevin. So, it's time for us to talk about the future and what it holds using all these exciting AI models. So, the first topic that I'm going to talk about is banner automation. Now, you might be thinking we talked about and actually saw a demo of images being resized using our lifestyle automation for banners. So, what exactly is banner automation? So, this is basically taking the banner generation way further by using text elements, call to action, logo, and resizing not just the images with the background but also adding all the elements that go into producing a banner, whether you're using this for
a paid campaign on Google, Facebook, Instagram, or you're using it for social media. And this would make the process even easier. So, with the banner automation, you would have the ability to choose object markers, people, scenes, or auto mode, and choose the type of detection you're after, or let us figure out what is the area of interest. And then, set up as many banner settings as you like, based on the sizes. You would be able to control how many variants we generate for each size. So, for example, on the right-hand side, for a skyscraper size, we could generate a zoomed-out and a zoomed-in variant. And you can see here that Crop.photo logo was placed at the top by our AI, the description in the middle, and then the call to action. And here is a different variant of that. So, with the banner automation, we would typically generate three or four variants because, while AI can do a lot of magic, it cannot replace the human's ability to discern what would really look good to their eye. So, to solve that problem, there will be multiple variants that would be generated. You would have the ability to hit a reload button and generate even more variants if the background that is being extended... So, in this example, you can see it added a bunch of left and right pixels to outpaint this image. If they're not to your liking, there will be the ability to then generate different types of backgrounds, with or without prompts. So, that's the banner ad or banner generation automation that will be coming out very shortly, in the next quarter or so.
**29:25**
Then, moving on, another exciting tool that we are looking at building is a listing analyzer. This is targeted at online marketplaces, whether you are selling on Amazon, Macy's, eBay. The idea is to use the images with info or without info that you put on a listing. So, let's say you are selling something on Amazon; you have six or seven images, lifestyle images, infographics, close-up shots, etc. So, we are going to parse all these images through OCR on the text and use the various language models, LLM models, to make sense of the text, to see if it violates the policies of that marketplace, or if your images and infographics stack up against the bestseller. So, there'll be built-in comparative analysis where we can look up a particular product category. Let's say you're selling a shirt, and so we could go and look up the best sellers on Amazon for that category and see how they frame their images versus yours and generate a report for you. And finally, connected with Crop.photo's smart cropping capabilities to auto-crop these images to increase the image score. And then finally, a brand voice checker. So, this is a tool that would allow any brand that is producing imagery for various channels, any outbound communication that needs to ensure the brand voice is honored in the produced images, or if there is text on those images, then it's not violating any of the brand rules. So, this could be technical specifications of the images or very specific social points that need to be identified. So, for example, maybe a brand says that there should not be any children holding a device in an ad, or a person should not be eating and looking at a device. So, this is going to use many different AI technologies that AWS offers, including what Kevin was talking about, Bedrock, to do analysis on the images and then also many different LLM models to provide insights into not just the text but also the images. So, that as a brand, you can ask pertinent questions like, "I have these 10 rules, like there should not be any violence in the images, or there should not be any children below six years old," etc. And it would enforce that, generate a report for you if the brand guidelines were being violated.
**32:23**
So, that brings us to the end of this webinar. And in conclusion, as you can see from the demo that we showed, AI currently is perfect for low to mid-end editing or retouching tasks, and especially when you're doing this at scale, when you're dealing with hundreds of images. And most of our customers see a huge productivity boost, two to three times, and saving off hours and hours of creative time. And the roadmap is constantly evolving; you can check our roadmap on the new automations and AI features that we are adding. So, with that, we will take up any questions that you might have.
**33:18**
Okay, so there's a question here: How many images can your tool handle at once? So, at this point, we basically can scale to roughly 2,000 images. You can upload simultaneously in batches of 2K and then trigger a crop. And so far, we think that that is sufficient for most use cases.
But we run on the Amazon elastic infrastructure, so if you need even more than 2,000 images to be processed, there are ways we can scale up to a greater number of images.
**34:03**
There's another question: Does it have Okta integration? So, we do have single sign-on support. Right now, we offer it through Google Single Sign-On. For our Enterprise Edition, we can offer single sign-on with other identity providers, including Okta. Please talk to us if you have a specific integration in mind.
**34:24**
So, another question is: How does the integration of Crop.photo and AWS Recognition improve the accuracy of crops? Maybe Kevin, you might want to talk about Recognition and accuracy a little bit.
**34:37**
Yeah, for sure. So, I think the value of an approach that uses a pre-trained model like Recognition is that they have been trained on tens or hundreds of thousands or even more images of a certain kind. And so, if you look at Amazon Recognition's homepage, you can see some documentation. We published something like a scorecard that talks about the service, the information behind the service for Recognition. You know, talk a little bit about accuracy in this space, but you know, what you'll see is that the quality of crop that you'll get from Crop.photo, because you're using a pre-trained machine learning model that's trained on so many images, will give you highly repeatable and highly accurate crops because it can very consistently identify those points across a wide demographic.
**35:29**
Yeah, and to add to that, that is really the hallmark of Crop.photo service, to use models that are highly predictable in terms of recognizing specific features. And the other thing to note is when you do bulk or batch image processing, often you have shot your images against similar backgrounds, so the patterns are pretty much set. And you know, when you do a free trial of our service, you can see for yourself how well it's working with a couple of images, and then you can be pretty much sure that if you scale from 2 to 100 or a thousand, you will see similar results with the various analyses that we do.
**36:05**
All right, the next question is: When doing batch crop of images, can you add custom images to use as a background? Absolutely. I actually did use a custom background when I was showing you the crop, but you can input as many custom backgrounds, and the tool will remember them, so you don't have to upload them again and again. And when you do the next automation, you can apply a different background. A lot of our customers, who are especially doing volume photography, like school yearbooks, etc., use that feature. In sports, that's used a lot to put like a sponsor logo in the background and then crop the images.
**37:28**
Okay, another question about security: What about data security and privacy if we upload sensitive images, such as new products that are not yet public? Yeah, so there's a lot to talk about security and privacy. Maybe, Kevin, you can start here, and then I can add what Crop.photo does specifically to secure customers' content.
**37:52**
Yeah, for sure. So, there are two kinds of services in the machine learning and AI environment at AWS: stateless and stateful. Stateful services are where we need to store some data in order to do what you're asking us to do, and stateless services are where they're purely transactional. And so, Amazon Recognition, as used by Crop.photo, is a completely stateless environment. So, an image is passed to AWS, or part of an image that's anonymized. We don't know who it's from; we don't know anything about it. The AI operates on the image; JSON is returned to Crop.photo to do whatever they need to do with it. And because Crop.photo has opted out of any of that data being used to train our models, there's absolutely nothing retained by AWS or Amazon. So, it's a completely stateless and, if you will, forgettable transaction.
**38:43**
Thank you, Kevin. So, to add to that, Crop.photo is also SOC 2 compliant. So, those of you who are in the US probably are very well aware of SOC 2 compliance, but it's probably one of the highest certifications you can get for a service to ensure privacy and security. And there are a lot of customers, especially in the automotive industry, that are using us. They are cropping models of cars that have not been released, and you can understand how sensitive that is. So, there are other certifications, like TISAX, that AWS provides, and Crop.photo is able to support those. None of your data is visible to anyone at Crop.photo; it's all encrypted. It's literally like a zero-knowledge system.
**39:37**
Okay, let's go on to the next question: In terms of scalability, how do you scale during peak e-commerce sales periods? That's actually a great question. Basically, like I've been saying, we use the AWS infrastructure, and one of the biggest benefits of using that is the Elastic Compute Service that we use, and some of these services will horizontally scale based on demand. So, if a customer comes in and is trying to do 2,000 background removes or generative AI, we could spin up another server automatically. If two customers show up and they need to do similar-sized loads, it's literally infinitely scalable horizontally. And like I said, at this point, there are over 5,000 users on this platform, and there isn't any slowdown.
**40:24**
Okay, and probably take a few more questions: Will your banner creation functionality include variable image and text generation? Yeah, so with the banner creation, it will work either with a single image, generating multiple different variants, or you would be able to input multiple different images and specify the automation settings, like all the sizes that you want, and it'll target all of those images together. So, if you had a batch of, let's say, 10 products that you wanted to put on social media, you wouldn't have to repeat this 10 times. You just input the 10 images, the generative AI prompts; you can set it to auto, and it'll automatically expand that. Now, as far as text generation is concerned, right now, in the roadmap, we're not looking at doing text generation automatically. That's something in the future to consider, to completely automate the text part of a banner.
**41:37**
All right, so I think that brings us to the end of this webinar. And just a word about a promo: So, when we launched this webinar, we had said that you can qualify for free credits. So, if you look at this webinar and can copy this code: AWS CROP 13, so you can sign up for a free trial and just message us the code or book a meeting with us using this URL. And if you mention this code, we will be able to give you those free credits. So, I want to thank Kevin for joining this webinar today, and thank you all, and have a great day. We will be sharing a recording of this webinar with all the attendees, and those who could not join will also receive a link to view the webinar.
---
