Canva Text to Image: The ULTIMATE Prompt Guide

who can tell me what the difference is between this photo of a young boy with glasses and this one or even this one the answer is simple The Prompt I spent a couple of days researching how to get the most out of text to image canvas AI image generator and I can say for sure that what I discovered helped me get way better pictures so yeah I think I cracked the code and in this video I'm gonna spit it all out for you and teach you everything I learned about how to prompt text to image let's Dive In all right guys let's start at the beginning and see what is text to image so text to image is canvas version of mid-journey or dial e but it's baked in the canva experience now when you look at this from inside the editor in order to find it you have to locate the little apps tab right here on your object panel click on it and it should also be the first one that you see here this feature so if you click here you can start prompting by typing a description in this box right here you can see some examples and yeah once you have a prompt you can select one of the Styles so if you click on Co you see you have photography digital art and Fine Art so different kinds of stars are predefined for you and the last thing you can do is select the aspect ratio right here so you could have square landscape or portrait so that is in a nutshell what text to image is one thing I realized while doing my research about text to image is that if I want to understand how to best write my prompts I need to know what under the hood meaning what is the underlying technology canva relies on for text to image what powers it because canva is not an AI company canvas a design company so they probably rely on somebody else's API AKA technology to bring us this text to image app so I did a little bit of digging around and easily discovered that the technology behind text to image is actually stable diffusion which is one of the most famous AI image generator model by a company called stability AI so what canva did is that they made a deal with stability AI in order to be able to use stable diffusion inside canva and for safety reasons canva added a couple of security layers or filters on top of the base model so that people cannot prompt for things that are dangerous or obscene or you know unappropriate and finally to improve the overall experience canva came up with the predefined styles so that you don't have to specify that you want a photo or a concept art or a painting you can just click on one of their predefined Styles and avoid the hassle of including that into your text prompt all right we are going to get into the crunchy part and I will show you what I learned about better prompting in a second but before we get started I believe there are a couple of things you should know and the first one is that text to image will work way better when you prompt in English now English might not be your native language but try to prompt in English for the best result the second thing is that text to image is available to all users So Pro and free but and there is a but you will be limited to 100 images per day and 100 images is about 25 prompts because for each prompt you will get four result so 4 time 25 is 100 images generated with text to image so be mindful of that because that limit is easily reached the third thing you should know is that it works better when you talk to it as a human versus talking to it as a computer program okay so don't try to create two elaborated prompts sometimes simple short prompts will work very well so just talk to it as you would to a friend and then the fourth thing which is probably the most controversial of all is the copyright issue can I use stuff that I generated with text to image for commercial use I just happened to talk to a friend of mine who is a copyright lawyer yesterday and she didn't know as of today experts don't know what to do with generative AI in terms of intellectual property so if you want to be safe just don't use them for commercial use right now and wait for the dust to settle in the legislators to come up with some sort of regulations what you really shouldn't do is ask me that question in the comments because I don't know all right now that we get the formalities out of the way let's see how to best prompt text to image so how are we going to do this well based on my research I came up with two simple steps to really improve the quality of your prompts and therefore the quality of your outcomes with text to image the first step that we are going to explore right now is a specific prompt library that I discovered with a bunch of stable diffusion ready to use prompts so we are going to dive into this right now and then the second step we are going to explore a bit later in the video is a Formula that has been working very well for me and my prompts but let's start with the prompt Library I was telling you about the website is called Lexica and it is a stable diffusion search engine all right so here you will find a bunch of different images that have been generated with stable diffusion and the beauty of it for example if you like this rabbit on a motorbike right here the beauty of it if when you click here you should see the entire prompt that the person has been using to generate that image so from here you can easily copy that prompt with this button right here so you can do that then come back to Canvas text to image and paste that prompt you will not get exactly the same image because this process is pretty much randomized but it gives us a good idea of what people are using in their prompts to generate specific images so if you like something like this for example you have the entire prompt right here okay so let's try one let's copy this prompt okay so this is a lost astronaut in Jupiter epic scene displaying in the background the moon and other planets and stars and Killian and vibrant colors Dynamic lighting blah blah blah there's a bunch of things then there is some extra information that I thought I would cover it says trending on Art station so art station is another website where people like to show their digital images and Creations and then it says art by this Lao beskin's key probably pronounced that wrong so that defines the style of the picture Okay so this guy is an artist who creates digital art so by giving the AI the name of a specific artist it gives it Direction it gives it a specific style to emulate all right so I copied this style and now I want to come back to canva and copy that prompt right here in my text to image prom box I'm not gonna change any of the styles for now I'm just going to go for the square format and create my image to see if this comes even close to what I saw on lexicon so these are the results that have been generated by text to image let's click on a couple of them so this one right here kind of like the same mood and let's use this other one right here so these two have been generated by text to image I would say they are not as good as what I saw on Lexica but they're not bad either let me upload the one from Lexica right here there you go so yeah if I compare the original one this one with these two other ones they are pretty similar so browsing through lexicon and the thousands of images that you can find there to find an image that you like and then simply copy The Prompt and paste that into canvas text to image is a very simple way to generate stuff that you know will look good because it already looks good on lexicon okay a couple of more things I want to show you on Mexico and I'm going to try another prompt to show you kind of the variety of what you can do if you look under the image the main image that is selected right here you should see different very ends of the same image so these are the other images that have been generated by stable diffusion the first one being usually the best one the one that the person who prompted selected out of the four options so usually the first one will always be the best one something else is that you will also see a negative prompt most of the time just don't worry about this one too much because canvas text to image doesn't have a field for negative prompt while stable diffusion does have that feel alright let's close this one and come back up here because I want to show you a couple more things about this stable diffusion search engine because that really is what Lexica is all about you have filters right here and in the filter you can decide to choose either Lexica aperture or stable diffusion 1.5 so both lexical aperture and stable diffusion 1.5 are two different image generation mothers Lexica aperture being the newest one so it should run render the best results and I'm not completely sure which of these two mothers canvas text to image is relying on so I couldn't find that anywhere if you know let me know in the comment section but for now what I want to show you is that when you click on that filter button Lexica aperture is the default button and will give you some better result in my opinion let me just show you what you get from the stable diffusion 1.5 model you see the photos are not exactly as good so I would definitely recommend you use Lexica aperture in terms of browsing Lexica so that you can find the best possible images alright let's find another example before we move on let's find a cute photo maybe something more in the style of a photo I can also search right here so let's say portrait of a young boy uh smiling let's see what we can find okay a bunch of different portraits of a young boy smiling yeah let's try this one right here it says a kid sad old color oil painting realistic okay this seems like a good prompt I'm going to copy this prompt come back to canva add a page here come to my text to image and simply going to paste that prompt gotam at any I don't know gottam Irani so I might just look it up quickly Gotham Alan is an Indian billionaire industrialist who is the founder and chairman of the adani group all right so it's just some rich Indian dude okay so we have the original portrait here let's see if text to image can generate a similar portrait of Gotan create and you will notice that I didn't use any of the filters I just want to see how it works when I prompt kind of like without any other information so I do have some cool portraits here I really like this one this one is nice and I really like this one as well so these two are pretty cool let me grab the original one to make a side-by-side comparation so there you go it is uploaded and yeah we do have something pretty similar in terms of the texture in terms of the style oil painting text to image recognize that this prompt worked very well so oil paint not even oil painting it's just oil paint and I get a pretty similar result all right I don't have the Bindi right there for gotem but it does look pretty similar in terms of the kind of results that I achieved okay so you see this technique works very well before I jump to the second way to improve your prompts I want to show you a couple more images I was able to generate following the exact same strategy so these two ones right here about Mark Zuckerberg I found based on this other photo from Lexica so I grabbed the prompt came back to text to image put it there and these are the two images that it generated so again the original photo and the two ones that were generated similarly I found these vintage photo portrait of Elon Musk on Lexica with this prom right here so I copied that prompt pasted it into text to image and these are the two images that it generated pretty mind-blowing in my opinion Union like Elon Musk is really recognizable on these photos and yeah when we look at the original one this one right here they're pretty similar so again I was very satisfied with this result and you know me I'm a big Snicker head so I also had to try generated like high resolution images of sneakers and specific mothers of sneakers and these are the two images that canva was able to generate for me of the Air Max 1 model so this one and this one right here they're not perfect but it's pretty darn close alright so this is the first way you can really improve your prompt in text to image is simply to find inspiration on lexicon now let's move on to the second way you can generate better images with text to image [Music] in this second way of generating better prompts and therefore better outcomes for your text to image we are going to lean on AI copy text generator such as Chan GPT or even canvas magic right and ask them to generate our stable diffusion prompts for us so this is just a smart way to use AI to create a better way to communicate with another AI tool and this idea by all means is not something I invented I need to give credit where credit is due and I learned that trick from my friend Matt wolf Matt's also a YouTuber his channel is called Matt wolf and he talks about Ai and all the latest AI tools his channel has been exploding recently because everybody wants to know about AI he also has a cool website called Future tools where he organizes and reviews all of these AI tools so Iris recently had the chance to interview Matt for an episode of pro tips which is not yet published on this channel so if this is something you are interested in watching make sure you subscribe to the channel so you don't miss that video when it will be published in a couple of days all right so back to the tutorial Matt is actually the person I learned this tip of using chat GPT to generate stable diffusion prompts and yeah I think it was in this video right here it says epic chat GPT prompts The Playbook so yeah if you are interested I will have a link in the description to this video of Matt but yeah so I learned that trick from him what we are going to do we are not going to be using chat GPT to do the same trick we are going to stay within the canva ecosystem and use canvas magic right because it works just as well alright so first let me introduce you to the formula that I learned from Matt so the formula goes like this here's a formula for a stable diffusion prompt so an image of adjective subject doing action all of that between brackets creative lighting style so that gives us some info about the lighting style detail realistic HD trending on Art station remember what we saw on Lexica that tag that was there in the style of famous artist one famous artist 2 and famous artist three so what we need to know everything that is here in between these brackets will be kind of like randomly generated by the AI okay so they will know what's in between bracket you can have fun with this and play with different alternative different options so this first part of the prompt is what Matt is using to kind of train chat GPT and then it will ask something like do you understand the formula child GPT will say yes I understand the formula and then it would come up with the second part of the prompt which is the actual prompt and it will say okay then write five stable diffusion prompts using the given form when the subject being XYZ so XYZ being the subject you want to generate the photo or the image about all right so let's try this with different kinds of subjects using magic right to see what kind of prompts it can come up with and then we will use these prompts and run that in canvas text to image all right I am here in my test document let's open the canva assistant let's find magic right right here magic right is a pro feature but you will have the chance to try it 25 times for free but if you have exhausted all of these free tries you can one subscribe to canva Pro or two use chat GPT to do the same thing it will work just the same but here since I want to show you guys how to do everything from the canva ecosystem I'm gonna use magic right okay magic right so this is my prompt here is a formula for stable diffusion image prompt okay I changed the formula slightly to kind of adjusted to what I want here I want a photo as close to a real photo as possible so this is how I changed it a photorealistic image of and then I let the adjective subject doing action creative lighting all of this I let that as such in the style of and then instead of artist one artist Two artist three what I did is that I did a quick Google search for famous portrait photographers and I grabbed these names of famous portrait photographers and I paste them here right there as the inspirational style for this outcome okay write five stable diffusion prompts using the given formula with the subject being a young African-American girl so this is my prompt you will see I let some of the random stuff going on but in terms of the artistic Direction I gave it the names of famous portrait photographers alright so let's see how magic right can handle this formula and generate my prompt so here are my five prompts let me make that bigger for you guys gonna bump it up to 25 okay so this should be now readable by everyone a photorealistic image of a Serene young African-American girl gazing into the distance lit by a dramatic Sunset detail realistic HD trending on Art station in style or Richard avidan Yusuf karsh and Helmut Newton okay so these are the three photographers alright so what is going to change here is pretty much the subject and the action and the style of lighting okay so I'm gonna highlight this for you and we can choose together which one we prefer okay a Syrian young African-American girl gazing into the distance then we have a confident young African-American girl dancing in the rain okay then we have a playful young African-American girl blowing bubbles lit by Whimsical colorful blah blah blah okay so blowing bubbles a thoughtful young African-American girl reading a book okay so let's highlight this and then a fierce Young African-American girl boxing so these are my five options right here I think I'm gonna go with the first one and then we'll see where we go from here so I'm gonna copy this entire prompt bring it to the prompt box in text to image right here I'm gonna leave it on none in terms of the filters come to my white page and create this photo see what happens alright I have some high quality stuff right here let's click on the first one right here this is really really nice like this photo is super realistic indeed I like how she is indeed gazing into the distance in a thoughtful way this one is also cool yep like everything is well proportionate on this photo like there's no major like bug or extra ear or second nose or anything like that this is good um this one also good though the subject is much smaller so my preferred one is really this one right here so yeah there you go guys we see that the outcome from my prompt is a really high quality because I gave it some inspiration all right so before I move on to the next problem let me show you what text to image generated for the exact same prompt yesterday when I was prepping for this tutorial I was able to generate this one image right there for the exact same prompt so yeah the outcomes were pretty mind-blowing in my opinion let me show you a couple of other photos that I was able to generate using similar kind of prompts just changing slightly the subject I was using or maybe the name of the famous photographers I was able to generate this Indian older man wearing a turban and I really like this one I generated this close-up of the face of a Nation woman smiling using the same technique and this one right here was generated by using the same prompt of this older Indian guy wearing a turban but this time I changed the style being none to vibrant so when I use the vibrant style this is the image that it generated for me all right now that you know how the formula works now that you know how to generate your prompts with magic right let me show you how to go one step further and really find the right references that will give text to image the right direction when it generates your images and the way we are going to do this is that we are going to continue using the AI to generate lists of different artists famous in different art categories the first one being photography so let's use canvas magic right again or could be chat GPT to generate a list of famous photographers who have won prizes okay so I'm gonna copy this prompt right here open canvas magic right okay there you go and simply prompt generate a list of famous photographers who have won prizes give me their names only in a bullet list okay so canvas should do that there you go generated my list Hanson Adams Diane Arbus Henry Cartier bresen Robert Kappa etc etc etc so I could take one randomly Steve McCurry so I'm just quickly double check with Google who is Steve McCurry all right there's Steve McCary American photographer all right so maybe we can see some of his art and there you go we have here A bunch of different photos from Steve McCurry okay I see now the famous photo of this girl right here this one has been on the cover of so many different magazines so this shows me that the info is legit I have now a list of 10 famous photographer and you can go even more granular you can ask for portrait photographer you can ask for children's photographer you can ask for abstract photographer you can ask for photographer who were published in specific magazine so I have come up with this list right here of portrait photographers I will have a link to a document with all of the fruits of that research right here that I'm showing you on screen in the description of the video that would be a template link to this canva presentation so you can use that link in description and have access to all of these resources so what I did I did that not only for photographers but I also did it for famous graffiti artists who have been mentioned in mainstream media and that is what magic Wright came up with Banksy Keith Herring John Michelle basquia Shepard Ferry Invader as gmails etc etc etc alright so let's try to generate a prompt this time with this list of graffiti artists all right and I'm gonna be using chat GPT to just show you the workflow using a different text generator so I pasted my formula here is a formula for a stable diffusion image prompt what I change is this I want a graffiti of okay so instead of an image I say a graffiti of and then just let everything random and yes in the style of and instead of having three different artists I only gave it one so it should be more precise and more faithful to the style of this specific artist and the artist I chose is Keith Herring okay so do you understand the formula I'm gonna run the prompt should say yes okay so now I will paste the second part of the prompt write me a five stable diffusion prompts using the given formula with the subject being and then I'm going to change the subject I'm going to write children holding hand children holding hand okay let's run this prompt and see what's going on so a graffiti of Joyful children holding hands surrounded by warm golden lighting with a Whimsical and playful style featuring intricate details and realistic texture and HD quality trending on Art station in the style of Keith Haring so yep children holding hands under soft glowing light okay so the lighting style is going to be different I'm going to choose one randomly let's take this one here mysterious and Atmospheric lighting with mystical and fantastic style maybe not this one because Heath Herring is not about all of that dramatic and Moody lighting cinematic let's take the first one right here warm golden lighting okay so I'm gonna copy that come back to my text to image I'm going to paste that here alright so all of the prompt right here I'm gonna leave it on none for now and generate it to see what it comes up with is it going to look like Keith Haring yes the first one I see here is yeah pretty much Keith Herring so yeah I like that so the graffitis generated were these ones so we have this we have this that is pretty Keith Herring as well in my opinion we have this so all of these are pretty good actually and pretty faithful to my original prompt so yeah that was just to show you it works with different styles let me see if I can even refine this further by instead of leaving the filter on none if I post it as a concept art what would happen if I use that filter instead of no filter and compare the new outcomes with the previous one there we go so we have the fourth one my favorite of all of these probably this one right here yeah this one is really cool we can see the texture of the wall let me move these out of the way and yeah so this one we can see the texture of the wall we can make it bigger to really see we see a door so we can see this is a graffiti let me delete this to have this full screen this is actually very good because we can see it's the wall of a house and this has been painting So if we make this full screen we could add more elements to this we could have people walking in front of that wall etc etc and talking about graffitis I wanted to show you two other artwork I was able to generate using similar formula so this one right here of a city skyline and this other one asking for colorful City skyline at Sunset so these two were generated I was pretty happy about them so yeah showed you how to generate a list of graffiti artists in the document that I will leave in the description for you I will also have some caricature illustrators so all of these guys is right here I also left the prompt that I used in canvas magic right so you can kind of get inspired by this prompt and generate your very own prompt about some others type of art or other category of artists whether is it I play around with yes some famous 3D generating softwares because I figured not only the artist will influence the outcome but also the tools used to generate these outcomes okay so in terms of the softwares it generated this kind of list Autodesk 3ds Max blender Cinema 4ds ketchup Maya etc etc well that is it I ask I ask it to generate a list of famous nft Collections and there you go five nft collections that were very famous the crypto punks the board app the art blocks pudgy Penguins and the world of women and so I use my formula in canvas magic right and ask it to generate five stable diffusion prompts of a penguin eating a cookie in the style of the punji Penguins the famous nft Collections and these are the two best ones that it came up with so this was generated by canvas magic right just using the right prompt so I think I'm going to leave it here for you guys I think you get the point you need to one use the formula to ask the AI tool of your choice to generate your prompts for you and three do your research about famous artists famous art categories famous techniques famous softwares used to pretty much achieve these type of art forms and then if you mix all of that you will get the absolute best result out of canvas text to image right let's wrap this up with three important takeaways number one the mother the image generation model the AI model is trained on photos taken from the internet this data set so feedback these references to the model to generate the best outcome alright so that's exactly what I was doing when I was researching famous photographers famous painters famous graffiti artist because it has been trained on images with tags like this is Banksy this is Van Gogh this is Picasso it knows that and therefore if you use these keywords if you use these references and feed them back to the model it will generate the best outcomes for you number two leverage magic right and chat GPT to write your prompts for you because who is more capable of talking to a robot than another robot so yeah ask chat GPT or canvas magic right or any other AI copy generator to write the prompt for you using the formula I gave you again credit goes to Matt and third grab our prompt guide that will be listed in the description this will be a template link to this deck I just showed you so by using that you will have all of the references feel free to build upon that maybe share it in our Facebook community so that everybody can benefit from your more research this was my research I hope you enjoyed it thank you for watching until the end you are an absolute Legend if you're still with me at this time I'm Gonna Leave You with our playlist of the other video and tutorials I created about the other AI tools and yeah it's right there and it's full of gems like this video so

Channel: Design with Canva

Views: 76,434

Keywords: text to image prompt, artificial intelligence, ai, generative tools, openai, magic write, Canva, design, Text to Image, prompt, guide, graphic design, visual content, social media, marketing, branding, online presence, creativity, visual communication, visual storytelling, content creation, generative art, ai prompting, ai prompt tips, best AI art prompt, AI art, pr

Id: 1fhhFuup1HY

Length: 31min 27sec (1887 seconds)

Published: Thu Apr 13 2023