How I made an illustrated story book using ai

so I wanted to have a little bit of fun this weekend just trying to use chat GPT mid-journey and some other tools just to try and put together a little bit of a kids story I wanted to do a kids story just because I thought it might be quite fun and you get quite a lot of creative license when it comes to imagery and and what things look like also I thought just doing a story in general might present a few design challenges that might be helpful to try and overcome uh using some of this technology just to push it and see what its limitations are so if you're interested in sticking around to see how that goes then let's check out the video so jumping over to Chachi PT then I asked chat TPT to create a story about a little boy and a little girl It produced this story about two kids called Jack and Lily who essentially go exploring into the woods and find a little bird who's been hurt they then take the poor little bird to the doctors fix them all up and bring him back and the moral of the story is to to kind of do unto others and be kind and and the sense of good feeling that you get from that so that's great churchy PT has given us a a few characters to work with we've got a bit of a theme for the story so we can jump over into mid-journey and start trying to make it happen so jumping into our character creation then in mid-journing I wanted to start it off kind of simple so I just punched in a quick prompt to create a character model for a cute little boy and a cute little girl I ended up producing me a few decent examples but most of them weren't body shots I picked out the ones that were body shots cut away the background and then fed them back into mid-journey if you don't know that you can do this then in mid-journey and the thread in Discord you can add in images you click on the image and click copy URL then you can post that back into your prompt what that's going to do is mid journey is going to analyze that image and use that as part of this generation for how it creates a resulting image the images I've got were a slightly different style to each other so I ended up running them through Photoshop pulling them out the background putting them back through in the way that I just explained and producing more character models in the end simple prompts work just asking for a cute little girl and a cute little boy full body character models along with those two images it ended up producing few characters that I was quite happy with so I ended up selecting those as Jack and Lily and taking them over to photoshop to do the first part of the processing right and the real processing we obviously needed a background scene um so going by the first scene of the story where they talk about growing up in a little village I wanted to kind of generate that and try and create something that with a similar aesthetic to Jack and Lily's character models I started by trying to feed in the character models in a similar way that I use the character models before but obviously this led to generating a lot of images that had faces in um and and bodies in obviously because we were using the character models to generate in the end again I ended up getting away with just using some very basic prompts just asking it for a cartoon style cinematic establishing shot um of a city or a village and one of the reasons for doing this was to try to create a similar aesthetic across the background I figured that the AI would interpret the word cinematic as kind of a teal and orange kind of vibe and stick with that I've done a few iterations of this until I was happy with it selected one and then moved it over to photoshop in Photoshop now then and we're just playing around trying to fit the characters into the scene one of the nice things that ended up happening when we used the first original character models to generate our final character models was that we ended up with quite a similar light Direction on both of the characters which is really handy and also in some of these scenes there was very strong contrast in light so it was kind of easier to match the Shadows again this isn't perfect and I'm not shooting for like some perfect bit of work here really just exploring the art of the possible and where we can take it so the second scene I generated was quite similar to the next one I used the first image of the establishing shot and then asked for some Woodland it kind of ended up looking like it was just on the outskirts of the village so I thought I'd use this as a kind of transitional scene and one of the things I've learned when doing things like this with mid-journey and with this kind of AI generation is to not get too caught up on the on the on the Minor Details and just kind of focus on the macro also we can slightly adjust things and edit things in Photoshop another interesting note is obviously that now Photoshop has got all the neural filters and different Ai and machine learning tools for example photoshop's neural filter upscaling and also using the content aware fill tool just to expand some of these images to fill the frame one of the next stages then and one of the trickier things to do with mid-journey was to try to create our character models but in slightly different character poses um and with different facial expressions and different variations um when you feed the images back into mid-journey it will produce more character models but oftentimes there is quite significant variations and sometimes just small variations but always there will be some sort of variation instead of character and it can be like little things like a different colored shirt buttons in different places and things like that so again it's another one of those ones where at the moment you just kind of don't get too caught up on the Minor Details and just go for the general aesthetic of what you're trying to achieve for the next scenes it's generally just some more kind of wide establishing shots just kind of General scenery um it's not until we get to the part where we introduce the character of the bird that things get a little bit more tricky and the way I tried to approach this is I just wanted to generate a a nice image with the the final frame that I wanted with the bird in and then to extract that bird and kind of maneuver it around the frame slightly kind of in the way I did with the the original character models and then you'll see how we kind of come back to that scene later on in the in the story here we're just maximizing on the scene we've set up by punching in onto all the characters and giving them a little bit of dialogue just to kind of get through some of this narrative with uh fewer words so to do a bit of a conclusion then about this technology and this workflow I mean obviously there are a lot of limitations at the moment and this is just using mid-journey but if how quick the progression has been so far there's anything to go by then I can only imagine we're going to see this come on heaps and Bounds some things I would like to see would be just some more built-in kind of prompt buttons just to save a little bit of time so we're kind of working through especially if like prompt way in also be great to have a little bit more control over the inputs having some sort of prompt way in on images would be fantastic especially when trying to create more consistent character models it would also open up a lot of scope for more experimentation in terms of AI in Creative Industries I could see it's only going to get kind of deeper and deeper from here there's a few other projects that are being worked on at the moment I know there's one by Runway they've actually just released a feature where you can actually upload images and teach the image generation certain characters or certain things that you wanted to know and so really Keen to have a look at that I'll probably make another video with some of the guys at the office just kind of going through that at quite a deep level just to see what it can achieve but I do feel like we're getting to a point now where this stuff could really start to bleed into so many different workflows one of the things I'm really looking at and what this is a kind of really dry run at is just building storyboards for some of the videos we're creating our business if you want to check out any more about that that's optic matter you can check us out at www.opticmata.com if you haven't tried mid-journey yet and you want to check it out I'll also Chuck a little Link in the description um also a link for optic matter down there as well for anything else I've spoken about in this video I'll also leave some links down there as well if you want any more content like this feel free to subscribe and I'll speak to you soon foreign [Music] foreign [Music]

Channel: SmudgeOfficialUK

Views: 38,982

Keywords:

Id: tBcwyxmXeYQ

Length: 8min 46sec (526 seconds)

Published: Sun Feb 19 2023