VEO3 in Canva: Cinematic AI Videos Made Simple

1. Class introduction: Google VO three, which is the most recent and powerful AI Video model on the market today, is now available in Canada. That's right. You heard that correctly. I'm going to go over all the details. We'll see who can access it, where to find it. I will actually demo it to show you how it works. And also share Tanu'sFmula to write the best prompts for the VO three model, which is Google's latest tech for AI Video Generation. Trust me, you don't want to miss this one. 2. Google VEO3 now available in Canva: I think this is a very important news because Canva could have gone 1,000 different ways for their video generation. It could have continued its journey with runway, like it kind of started with previous version of AI Video in Canada, which wasn't great. Could have gone with Klein, could have gone with Halo, could have gone with really many different models out there. But instead of that, it went with the best in class. Google V three is the superior AI video model right now, and is absolutely amazing that Canva chose to partner with them, offering us users what's best out there on the market. So I really appreciate that, and I can't wait to show you how it works. All right, guys, this is the Canva homepage. So let me show you how to access, create a video feature. So for this, you need to get into Canva AI. Two different ways to get to Canva AI from the homepage. First one is via this button right here. Should say new still in your account or maybe doesn't say new anymore. That's this button right here. So when you click here, you'll get to the Canva AI. Or the other option is simply to come to the left menu right here, the sidebar and choose Canva AI from here. You'll get to the and from there, you should see something like this. So a page where you have different types of AI generated resources. Look, you see you have code, image, design, video, right, et cetera. So create a video should be right here in your account. This feature is going to be a paid only feature. Generating video, especially with the best in class video model today costs a lot of money to give you an idea, depending on which version of VO three Canva is actually running. If it's the fast model, generating one video will cost approximately $1. If it's the highest model, it actually closest to $5 at today's date. These prices are bound to go down, but this is the state of how much it costs as of today. And this kind of gives the context to what's coming. Create a video. This button right here is going to be a paid feature only. And here, paid means Canva Pro, Canva Teams and Canva Enterprise. I haven't heard about Canva for education yet, so I would assume it's only for pro teams and enterprise. So this feature right here, let's click on it to see what it looks like. So the interface is so I have a few related buttons right there. I have style, aspect ratio, duration, framing and lighting, like for the image generation and all the other AI tools, I can use voice command. Can chat using my voice if I wanted to. And then if I scroll down a bit, I will have my recent chats so my recent AI chats with the Canva AI tool. And then when I scroll further down to the bottom of this page, I should see a few examples of videos that have been created with creative video. So if you click on one, it will open it and start generating that same video using the same prompt. So be careful because you have a limited amount of prompts every month. So if you click on one of these, you are already going to be using one of your credits. How do I well, because I made the mistake myself. I clicked on this video right here of this cat walking and looking at the camera, and I wanted to see what the prompt looks like. So I clicked on it. It started generated the video. See, I had to stop creating the video. I hope my credit was not wasted, and I'm going to tell you exactly how many credits each user gets in a second. But I wanted to show you first that if you click on one of these ties, it will burn a credit and start generating a video. So let's have a look at this prompt, okay? So clicked on that cat walking video, and this is the entire prompt, which is, as you would see, quite a robust so highly stylized to the illustration of a black cat with off white face and pas cut mid stride with elongated limbs, a. Okay. So do you see, it's a very detailed prompt covering different aspects of that video. So this shows us that the quality of your prompt is going to have a big importance here, when it comes to generating videos with Canva and stick around because I have a formula that my friend Tianyu came up with that will help you generate these high quality prompts. We're going to get to this in a second. Before we start prompting, I want to kind of go over the details here because I believe they are important. I want you to know exactly what you can expect from this tool. So first, it will generate a 16 by nine up to eight second video with synchronized audio, meaning that it's not only video that we generate. You have your video with relevant sound. So it could be sound effect, could be dialogue, that the characters will just speak in that video. It's not just video like in Sa with no sound. It's actually video plus synchronized sound. So this is great. In the 16 by nine aspect ratio for now, when we come back to the interface and we check the aspect ratio button right here, we see 16 by nine, and we also see that the nine by 16, so the mobile version of these videos is coming soon. So that is exciting. We can see that this is just, I would say, the first draft of this tool that Canva wanted to put out there. So this is good to know that this is going to continue to evolve. Do we know about this tool? Well, every prompt will generate one video, one video only per prompt, so use them carefully. And that video that will be generated with creative video in Canva can be used, can be further edited in the Canva video editor. So you can open that in Canva. I imagine this video will kind of loaded under my uploads tab video. I imagine that's where it's going to live, but we'll find out in a second. Next thing we know is that it's available for pro teams and enterprise users only. And remember, I told you it was quite expensive to use this model. Well, that kind of explains this or justifies this. As a paid user, you'll only get five credit per user per month. If you paid attention to the previous AI video model in Canva, it was up to 50, but the quality was really not this one is more expensive, but you get less video. So let's hope Canva kind of pushes this limit up a bit or a lot in the future. Also very important because I know a lot of you guys are watching from all sorts of different countries. This is only going to work when using your Canva account in the English language right now. Doesn't mean you can only generate English speaking videos. You can change the language spoken in your videos in the generated videos by just specifying speaking that language. But in order to have the feature in your Canva account, you need to set the language of your Canva account to English. Pay attention to this slide, maybe screenshot this slide because every point is kind of important. 3. How to write effective video prompts for VEO3: Alright, let's get into the weeds here. Let's start prompting. Let's start generating these videos. Just remember that in order to generate good AI videos, you will need very good prompts, right? And for me, the go to person, when it comes to writing better prom for either image generation or video generation is my good friend Tianyu Tanu and I have created multiple courses. There's one about DALE. There's one about image generation for HGPT. But Tanu is just an amazing teacher. I just knows so much about AI. Here is a slide to follow him on LinkedIn. I highly encourage you to scan this QR code, follow TNU if you are in AI video, AI image generation. Let me show you his profile. He recently became a top voice on LinkedIn because of the quality of his content. So definitely someone to follow. This is the course we recently created together, the CHI GPT Image generation. But there is one particular post I would like you to pay attention to. This one right here. Goes like this. So here we can see a little demo that TNU created with VO three. And in this post, he shares effective VO three prompts in that formula. Cascade stands for camera, ambience, subject, context, action, dialogue, emotion. Okay? So these are the secret ingredients. You should be including in your AI video prompts in order to get the best images. You can trust him. He has played with the tool with Vo three since day one. He always acts like that, so he just jumps on the new tool and just and prompts and figures out the best way to use them so he can create educational content about them. So he gives us an example here for one of the videos that were generated right here. So what I have done based on that, I have copied the cascade kind of framework, right here, or structure and the example that he gave us. 4. Using ChatGPT to build a storyboard with VEO3 prompts: And I came to HAGPT where I prompted the following. Help me come up with a series of five scenes to create a little AI video. I will then generate each of the scenes with Google VO three. I want my story divided into five scenes to be about a rubber duck announcing we can now create AI videos with Canva. So that's me kind of giving the high level story or goal of my storyboard. What we are trying to create really is a storyboard with each of the fives. Going to be using this prompt structure cascade to write each prompt. And that's where I pasted what Tanu is teaching us on his profile, right? So the cascade, what it stands for. And then the example, your task is to suggest me five scenes that would make my story come to life. Okay? So that's exactly what CAGBT did. Like, I gave me five scenes, okay, one, two, five, and then I basically asked Cha GPT to open this in Canvas mode, and I started to add a little bit of my input. Okay? So I read each of the scenes. I made a few tweaks here and there until I came up with these five scenes, all have dialogue. Okay. And yeah, so now they are ready to test into Canva. And again, this is me on the fly. I haven't tried this before shooting, so I hope it's going to work. Okay, so scene number one, it's about this rubber duck. So the duck awakens with an idea. So I'm going to copy this and see how it goes. Alright, so let's come back to Canva. Here. Let's choose 16 by nine and paste this. It's delete scene number one. So Duc awakens with an idea. Close up vlog style camera angle, early morning, warm sunrise lighting. Rubber duck with soft light blocking sleep mask sitting upright on bed, surrounded by scattered notes and scribbles, Doc quickly turns toward the camera, excitedly says, Oh, my feathers. This is incredible. Can't wait to tell the whole world. So this is pretty cool. 16 by nine duration. Let's go for 8 seconds. Framing. Let's go. I think I gave the right framing here. I said, close up. So there are really two ways to go about these prompts. You can try to build them by using the buttons. Or if you like me, are using a framework in HAGPT to kind of build your prompts, I think it's enough to just past the description. I kind of want to try it like so. So let's go and generate our first video. Let's see. 5. Generating our first video with VEO3 in Canva: To Okay, so there you go. It's thinking on it. This might take up to 2 minutes, so hang tight while I get your video ready, okay? So you see it's queuing up the digital director. There's always a few kind of, like, jokes going on right here, teaching pixels how to dance. So I'm not going to make you wait here for 2 minutes reading these jokes. I'm going to stop the video, come back when it's ready. Alright, about a minute and a half half passed, and this is what I get. I see a video here. When I hover over it, I can see there are some subtitles, some captions. The duck has been correctly generated. So let's click on it. Oh, my feathers. This is incredible. I can't wait to tell the whole world. Alright, so this is kind of good. Like, it's actually really good. The only thing not great are the captions, right? So there are some typos in the captions, but the voice is incredible and the movement, the video, everything. Did you catch the little sound effect when it just flips the eye mask up? Pay attention. Incredible. F. I can't wait to tell the whole world. So different layers of sound here. We have the soft music of the morning awakening. We have the sound effect when the duck flips the mask up, and then obviously there is the voice, which is pretty cool. So from here, I can do a few different actions, actually, two different things. I can download this video. I can use this in the Canva editor. I'm going to download the video. Alright? So it's going to be downloading Okay, I have it here, so I'll come back to this later. Perfect. I have my first video. This is great. So I guess from here, I could either follow up and further prompt to modify the same video. As I told you, I'm not sure exactly how many credits I will have as a tester, as a Canva expert. Do I have more than five? Do I have five? Just like everybody sure, and I really would like to generate as much of my story as possible. So not going to follow up on this one, but I believe it's possible to follow up on this first video to ask for changes. Like, I could ask for no captions, for example, but I will first focus on generating the other scenes of my story, okay? 6. Generating the other scenes of my story: I'm going to come back right here. Click on Create a video again and just populate with my second prompt, right here. So scene number two, this is the scene number two. Let's come back. I have the prompt. Let's paste it here. Okay, scene number two, so that is the duck preparing to share the news. Okay, let me read the entire prompt. So duck preparing to share the news. Mid shot, playful camera tilt. Right cozy home office with natural lighting. Rubber duck at mini duck size laptop, typing quickly. Desk is covered in creative sketches and colorful decor. The duck pauses dramatically raising it wings and says and says Canva just changed everything. AI video are finally here. Then I would add no captions. Alright, cool. I don't think I need to specify the aspect ratio. I'm just going to put it just to be sure and launch the second scene. So here again, Canva is letting me know it's on it, it's queued up in the digital director. So as this feature rolls out, we can expect things to kind of get slower. Like, as of today, me filming the video, I have early access to the tool, therefore, it seems to me that it's quite fast, but I'm not sure how fast it's going to be millions of people are going to be using the tool at the same time, especially in the first days. Alright, let's speed up the video and see what this video looks like. There you go. My second video is ready, and I'm ready to discover it. Canva just changed everything. AI videos are finally here. That is about right. I love the little wing typing on the keyboard, the sound effect. Canva just changed everything. AI videos are finally here. And I like that it didn't create any captions, just like I ask, and the light is incredible on this video. The rubber duck is consistent with the first one, so this is great. Let's download this. And let's add this and rename it. So this will be scene 02. Perfect. My little project is working very well right here, so I can come back. And just go on with prompt number three. So prompt number three, scene number three, the duck is on the set. Alright, let's paste it. Let's read it. So that's scene number three. Again, I don't think I need to edit my prompt like this. And again, I will say no captions. Okay, duck on set. Wide angle log style camera capturing DIY filming set, vibrant atmosphere, bright studio lights, rubber duck adjust bowtie in front of a green screen. There is a bathtub with soapy water on the set. The duck clears his throat theatrically and address camera. Ready for this? Creating videos is now easier than floating in a bathtub. That sounds cool. Let's generate. And for those of you who missed it, now that you've seen the power of Tanu's prompt formula, I'm going to bring back that QR code so you can really get there. Follow Tian u on LinkedIn and let him know Ronnie sent you to his LinkedIn profile. Alright, our third video is ready. Let's have a look. Ready for this? Creating videos is now easier than floating in a bathtub. Okay, so this time it didn't skip the captions. So yeah, I might come back to this afterwards. Like, after I have my five scenes secured, I will come back to this and try to yeah, try to get rid of the captions. But for now, it's me securing my scenes, right? I have three out of five scene number three. Perfect. Next one is scene number four, and I'm going to come back here. You should see also create video. So if you open your sidebar right here, you should see all of your different conversations. See, that's my video number one, scene number two, scene number three. Because I'm creating a new chat for every scene, it allows me to have them all listed there and come back to them if I need to. So let's go for scene number four and just a little bit of housekeeping hoops. I launched it without asking for no caption. So we'll see I believe this one is going to be the most difficult one because I'm asking the AI to kind of generate Canva AI video interface on a laptop. Over the duck's shoulder. This one is by far, in my opinion, the hardest one to generate. I'm kind of curious how the model will handle this one. It actually looks cool. Look at this Canva generated interface. Let's click on it. Just describe it. Click and boom, cinematic AI videos in seconds. That's pretty cool. No captions. That's a great little success right here. Okay, so let's rename that properly renamed scene number four. And now is the big moment. 7. Fine-tuning and adjusting generated scenes: Let's see if I still have a fifth credit. If I do, would be amazing. Okay, so let's come back. Let's go grab the last scene prompt right here, cascade prompt. Come back here and just go for it. Scene number five, the grand finale call to action. Alright. I want to have time to read it to you. This time. So grand finale, call to action. Front facing log camera, medium close up, zooming out, cheerful, bright window lighting, celebratory mood. Rubber duck in party hat and sunglasses, cofeti gently falling. The duck waves energetically and enthusiastically calls out. Go on. Try create a video in Canva, and let's make some splashy creations. No captions. Do Do not hard code any captions. Okay, let's be more precise. Let's go. Please generate my fifth seen. Yes. Seems to be working. Maybe I'm kind of off the hook as a tester, as an early user of the feature. We've been told that it's going to be five video per month per user. So maybe this one is my last one. We'll see, obviously, I will try to keep prompting. I'll try to get rid of the captions. Some of the videos I generated that have the captions to get a cleaner result when I put everything together. But yeah, I'm curious to see if I can have more than five. Alright, the video is ready. Let's click on it. Come on. Try create a video in Canva, and let's make some splashy creations. Ooh. Yeah, so this time. Come on. Try create a video in Canva. And let's make some splashy creations. I have some kind of Vietnamese looking text right here. Go on. Try create an video. So I would like that to go away, really. So I'm gonna download this just to have it, okay? I'm gonna rename it scene five, so rename n five. So I have it, right? But now, I would like this to go away. Can you remove the written text of the video? I don't want any captions in it. Let's go. Got it and make sure there aren't any captions in your video. Okay? So let's see if this works. If it works, it will be amazing. And you see this time, I just followed up in the same chat. Okay? So this is the chat of scene five. Alright, so I can see that it's not what I thought. Like, it didn't keep the exact same video. It recreated an entire new video. Let's see. Go on. Try create a video in Canva, and let's make some splashy creations. Okay, this is better. I mean, I like the other one, too, but at least this one doesn't have any caption. So I'm going to download this one, rename that scene five. No captions. Okay, so first video needs to be redone. Second one was good. Third one needs to be redone. Fourth. Okay, so I need to redo first and third. Okay, let's go find them and do the same thing. So let's come back to my main creative video page. Open the side menu. If it's not open like me, if it's collapsed, if you see it closed like this, click on the open menu Hamburger, and then find that chat. Okay. So the first scene, this one right here. Great. Can you remove all text from the video? I don't need any captions in it. Let's try. And then while it's doing this, I could try to do a third one, like the third video in parallel, okay? So that would be this one. Maybe I could just copy this line of prompt, come back here, paste the exact same one and lounge at the same time. So if these two work, I'll have all of my videos, and I'm happily seeing that I can continue. So maybe what's going on is that I can have five chats going on and one credit is one chat. Not sure. Maybe it's just that because it's not yet rolled out to everyone, they give us more credit to test, but I'm not completely sure. There you go. The first one is ready. Let's click on it. Okay, it's very different from my first video right here. Feathers. This is incredible. I can't wait to tell the whole world. Okay, so this didn't work. This video has nothing to do. Like, there's no there's a hand coming. It's very different. So my follow up prompt didn't work right here. Okay, so you know what? Let's try one last time. Coming back here, I'm just going to paste the entire prompt again. Do not generate any captions the video. Let's try to generate the entire thing again. And in the meantime, go want to see video number three here. Let's see. Okay, so again, like I lost the whole thing. So I'll just try again, same thing, repost my entire prompt and specify. Do not generate any captions for the video. Okay, relaunch that. And if that doesn't work, I will just have to find a way to work around the subtitles, the captions when I edit the videos. Alright. The first one is ready again. Let's see. Oh, my feathers. This is incredible. I can't wait to tell the whole world. Okay, so I love the little I mask there. It's better than the black one, but I still have the caption. So I'm gonna leave it here. Let's just have a quick look at the third one right here. Yeah, it is ready. Ready for this? Creating videos is now easier than floating in a bathtub. Okay, this one's not bad. I'll keep this one as well, because it has no captions. So this is scene number three, no captions. Okay, perfect. 8. Assembling the full video in Canva: So, as you can see, I have all my videos right here. All I need to do is to come back to Canva and start a video project. Okay? So here from the homepage, I can just go video. I'll create a landscape video like so, and then quite simply, I'll come to my upload videos, upload files, and just upload all of the video scenes I have generated. There will be all uploading here, and then I can start editing this little video right here And so let's see if I can make the text disappear. Okay, starting with scene number one, which is this one. Going to click on it, page two, three, four and five. Okay, let's add page two there. Page number three was this one right here, and then page four was perfect as well. And then page five. My favorite was still the first one. So now I just need to make this one set as the video background. Now we have the video kind of making sense between all of the scenes. It's really up to me to add transitions and additional stuff, but I would like to, get rid of this. Alright, so here for the first clip, an easy way to get rid of this would be to split my clip right here. So I will hit my S keep. I split it my clip. When I click on the second clip now, this here, what I can do is to simply stretch it. So I'm going to stretch this creating that zoom effect on my duck and making sure I don't see the text. And so there you go. Let me zoom on my timeline by clicking that little button right here. There you go. So I can't wait to tell the whole world. Canva just changed everything. AI videos are finally here. Creating. Okay, so here's going to be a bit more complex. So here you see, I have the same sound effect. I have the sound effect, and then I have the caption. So I do the same thing. I will cut right when the sound effect stops. Create. And then zoom in when the duck starts speaking. So here is where the subtitle or the caption starts. Again, I'm gonna split. So splashy creations. Okay, so maybe I could get rid of the image here, and I'm just adding something on top of my video right here. Okay? I'm gonna make this video transparent, boom, so I still have the voice. Make some splash. And I will bring the button. Actually, I have made a screenshot of that button. If I come to my upload images, you see, create a video, so I could bring in the button, position it at the middle of the page, and kind of have my background the same color as this button. So I'm using the color picker to pick the blue color here, so it kind of looks seamless. Let's play that again. Go on. Try create a video in Canva, and let's make some splashy creations. Okay, so that's not bad. I would probably need some sort of animations here where I click this button. Alright, guys, I worked a little bit more on that edit, and I'm now ready to show you the final outcome. Oh, my feathers. This is incredible. I can't wait to tell the whole world. Canva just changed everything. AI videos are finally here. Ready for this? Man, man. Creating videos is now easier than floating in a bathtub. Just describe it. Click and boom, cinematic AI videos in seconds. Go on. Try create a video in Canva, and let's make some splashy creations. When? 9. Fast forward 100 videos: Hey, guys. Again, what I suggest we do now is a little jump in time, let's fast forward, like a week or so. You know, I've been using VO three a lot. As a Canva verified expert, we were lucky enough to be given limited VO three credits in Canva so we could test the tool and generate educational content like this class for you guys. So I prompted a lot. I have run a bunch of different experiments in order to see if I could improve the character and scene and voice consistency between the different clips I generated with Veo three. So I'll show you the results of that and this improved prompt lecture. Also, I have discovered that we can generate video clips with Vio three, not only in Canva, but in Leonardo AI. So we will explore generating videos in Leonardo and see how the video generated compare to the ones generated in Canva. And then I will give you some tips and tricks to reduce the probability of having these pesky hard coded captions or generated VO three videos. So I found out I have talked to a lot of people in that community, and I watched a lot of tutorials myself, and I found a few very useful tips to not completely get rid of these hard coded captions. I don't think it's even possible as of today, but at least reduce considerably the occurrence of these hard coded captions in your VO three videos. And then when all of this is done, towards the end of the class, I will also show you how you can generate some new captions directly from Canva. So let's get back to Canva. Let's get back to the demos because I really want to share everything I've learned prompting these dozens of videos with VO. 10. Character and Scene Consistency in VEO3: The most common question I've seen on social media when it comes to VO three is, how can I make my characters consistent from one video generation to the next? And you have to understand that the model in itself, because it's primarily a text to video model, is not yet great at character consistency. It would be much better if we could upload an image that represents the starting frame, for example, and then run from that, like we do in ling for example, we have the first frame, so we can even upload multiple and the model will stick to that reference image in order to generate our video. It's not yet the case with Video three. The model is still relatively new or very new and therefore not there yet. And I know that we will get there because in Canva, we already have this plus add media button. You see? We can add an image as a reference. But when we do and try to create a video clip, it's not taken into consideration. So we'll get there, but it's not there yet. Well, the good news is that there is a way to achieve this character consistency going to be 100%, but I would say, as of today, it's around 80%, but it's good enough for the human attention span and, you know, the level of detail we actually do pay attention to as of today. So I have this project right here. I call this one the printing press, the Gutenberg project. It's a mini series I have created where I kind of, like, redo the historic moment where big innovation inventions were kind of discovered. So this one is the printing press by Gutenberg. See this is my character Gutenberg, in his workshop. And I'm going to show you this scene, one video. And this is scene two. So you see the character consistency between this scene and this scene is pretty good. We have kind of kept the main characteristics of this character. Next scene. So we have a slightly different color of leather apron here. This one is more blue. The first one was more black, like you can see, but these kind of details, they don't strike you because you don't pay attention to this level of detail, right, or the main like most people want. This is yet another scene. This one right here, you see, very consistent. So I have generated for this video four scenes. So I'm going to play you the entire video, and then I'm going to take you behind the scenes where I actually generated these videos, but also in chat GPT, where I started kind of scripting the entire concept. So let's first play the video. Before me? If you wanted a book, you had to copy it by hand. One letter at a time. Slow. Then I pulled up with this bad boy. Mass printing, baby. Books went viral before it was cool. Hit that like button if you've ever learned something from a book. 11. Fixing my prompt to improve consistency: So for me, creating a video with VO three always almost always starts in HGPT. I have created a project that I call VO three. So paid HGPT users will be able to create a project. Organizing these conversations, these chats into a project allow me a few different things that are quite useful. The main one being that I can input custom instructions will be taken into consideration for every new conversation I start within this specific project. So the custom instructions are here, and this is where I have kind of laid down my prompt structure, right? The famous cascade prompt structure, cascade stands for camera, ambience, subject, context. On dialogue and emotion cascade. That's the acronym. I added right here my custom instructions, the T for timeline because I realized when I create different scenes for the same video and I'm trying to achieve this character consistency between the different scenes, it is useful to have this element of time. For example, where does that scene number three, the printing press, whatever, where does it, like, stand compared to the other scenes we've just generated, right? So by adding this element of timeline, basically, including in your prompt where this scene takes place in the story will be use. So this is the cascade with a T or cascadete. I don't have a good acronym. Sorry about that. But also, I told you, I was going to add anchors, okay? So the anchors are these more robust character and ambience description that we will be creating, and then we will give them kind of a name or tag that we will be using in the cascade. So what are anchors? Well, it's just like a way to anchor the description of my character. So I would recommend you create a character anchor and you give it a tag. Okay. You can call it whatever you want. I'm going to show you an example in a second, and an ambience anchor. Give it another name. So these two things are detailed descriptions of your character and the ambience or the vibe or the aesthetic of a specific scene. Then you have your cascade prompt, the traditional one. Here in the same custom instructions, I give CHAT GPT an entire detail of how that will work. So it's always good. We call this one shot learning to kind of give HAGPT one shot, one example so that it can learn what kind of prompt I want to create. So we give it the prompt theory, and then example, so it has all the information it needs to understand where we are going. So this is my example. So the character anchor, I gave it the tag Gutenberg. So mid 15th century Joanas Gutenberg, about 50, broadow, keen green eyes, salt and pepper beard, shoulder length, wavy dark hair beneath a soft felt cap five foot, 8 ", solid frame, ink stained hands. Wears an off white linen shirt, dark leather jerkin, well worn tool belt, and a splattered press apron. Rhineland German accent, confident, baritone, quick wit, use the same TTS voice D throughout. So you see this line, use the same TTS voice ID throughout. This one line might just help me create that voice consistency throughout the entire project. If I prompt several scenes so different prompt, I generate different videos with the same character anchor. I will have at least tried to stay close to that voice that my character has in the first prompt or in the first. So I'm not sure 100% this works, but it did improve the consistency, as you've seen in the video I just showed you. So use the same TTS voice idea throughout. So that is my character anchor. I call this Gutenberg, okay? Second is the ambience anchor. I call it press Vives. Okay, so bright midmorning beams, spill through tall million windows, 5,500 K daylight modes of paper dust drift in the light, continuous wooden press creek, gentle metal type clicks, and a muted scent of wet ink and vellum. The scent, I don't know how used in my prompt Oak benches piled with movable type, freshly printed folios, and bound volumes to behind. Mintain this palette across all scenes unless specified otherwise. This is important. Maintain this palette across all scenes. So this color palette that aesthetic, that kind of filter that will be applied to the video needs to be repeated in every single scene of the video because this is really what will create that feeling of consistency from scene to scene. Alright, so these are my two anchors Press vibe and Gutenberg for my character. Let me show you how I ask JAGPT to kind of blend everything together. Okay? So scene 01, the setup. Again, this is still an example. You don't have to write all of this. You can ask JAGPT to write it for you. And this is what I did here. This is just the example I give in the custom instruction so it knows how to proceed from there on. So name of the scene, scene number one, the setup, okay? And then I have my cascade. As cadet because I have the timeline, the Scene one, the setup, first, the C camera, front facing log, medium shot. Ambience. Then we have the first tag here. We call press vibes, which is my ambience anchor. See, I gave it that tag. So warm indoors tone. So I don't need to repeat the entire thing here because it's already in my anchor. So that's the ambience subject. Obviously, we call our tag Gutenberg with parchment. He has something parchment in text, messy stacks of handwritten books, scrolls, loose pages. Action, gestures to the chaotic pies behind, so it kind of refers to that dialogue. Before me, if you wanted a book, you had to copy it by hand, one letter at a time. Slow emotion, mock frustration. Timeline, this is the opening. So this is just an example. Going to save these custom instructions. This will give ChaGPT a solid place to start with because everything I will write or create from now on in this project will be kind of guided by these custom instructions. So let me come back to my project right here. Let me go all the way to the top of this conversation and show you how I started. So this is my first prompt. Let's create the four scenes of a video that retraces the story of Johannes Gutenberg, the printing press. We will be using the anchor press casceded structure from the custom instructions. 12. Generating scenes in Canva with our improved VEO3 prompt: Your first task is to help me create one, the character anchor and two, the ambience anchor. For every new scene, pulling these anchors with the tag, only describe what changes. So camera, action, dialogue, emotion. Keep each scene under roughly 60 token. Concision helps the model focus. So basically, this 60 token is kind of the length of the prompt. I found that number by basically chatting with HGPTO three model, asking it, what's the best way to draft a prompt for VO three? And this number came Alright, so this is the high level story. This is me giving the high level story to Cha GPT in the prompt so that it knows roughly where we are going, okay? So Johannes Gutenberg, the printing press filming in front of a stack of paper or books before me, if you want to the book, blah, blah la. So that's the dialogue. So it greens, pans to the press, and then it says the following then I pull up with this bad boy, mass printing mic drop. So I changed that. I replaced every single piece of dialogue with an Imoge with something else. Because if you input Imoges into View three as a prompt, it's not going to work. So don't do that. On the other hand, use punctuation marks, use exclamation marks, question marks, three dots to create kind of a pause. This is very useful. This will guide hagiPT. So punctuation is very important. And then I have the following dialogues. Okay? Books went viral before it was cool, and then double tap if you ever learn something from a book. And again, some Imoges I think I changed double tap for hit that button. But yeah, so this is pretty much how it started. And if you're wondering how I got this high level story, I just ask Chat GPT in another chat to help me come up with my four scenes, and it gave me this high level story. So remember, I asked to start first task is to create the character and ambience anchors. That's exactly what it did here. So these two anchors, these are the same thing. I just radio because I used it as the example in the custom instructions. But yeah, now, give me scene one. So once I had the anchors in the same conversation, I started prompting for each of my I ask, now, give me scene one the setup, and this is what it gave me. It used the formula, the CascadetFmula, and it gave me the exact prompt. So what I did simply was to copy this, and I came to Canva AI. So this is Canva AI. Create a video. Let me go back to the specific project. So when you click on create a video clip, so by default, Canva, you will land here. From the homepage, you land here with no button selected. So select Create a video clip. And if you open that side panel, you should see all of your conversations. So my recommendation would be create a new conversation every time you start a new project. Okay? This is probably the best way to work. It doesn't mean just like HAGPT that it will remember everything from the conversation. Not sure about but it does help stay organized when you want to come back to your conversations. So Johannes Gutenberg, as you can see, what I've done here. So this is the first prompt I pasted. So the way you will prompt every time is that you will paste your two anchors. I know it's annoying to repeat yourself again and again and again. But for now, remember, I'm not sure that Canva remembers the previous so repeating the anchors is the only way I know to kind of stick to that character consistency as close as possible. So you will repeat these two anchors again and again, and then you will enter your new scene with the cascade framework. Okay? So that's exactly what I did here. And it generated this video right here. Let's play it. Before me. If you wanted a book, you had to copy it by hand. One letter at a time. Slow. Okay, so you see, this is correct. There's no captions or anything. So I was pretty happy about this video. I could have used that definitely. So I continued prompting. So I came back to HAGPT and said, Okay, now I need scene number two, the reveal. Make sure to add the timeline to the cascaded structure. So timeline is the last. I had the first cascade prompt without the tags, without the anchors because I asked ChaGPT to create that in a separate conversation. Let me show you the full thing I did here. So this is another conversation. You see, I'm using 40 for this one. I said, now, help me break down this video into scenes. So this what you see here, high level video idea. I also generated with HTPT. There's no cascade structure, no tags, no anything. It's just the high level idea of my four scenes. So I ask ChaiPT help me break down this video into scenes. And because I am working in my Video three projects that has the custom instructions, it used the structure from the custom instructions. So it gave me scene 124 with the cascade prompt. So this is what you see here. When I pasted that here and I'm working with the anchors, see? To try and create that character consistency. I ask it. Okay. This is my original prompt, my cascade prompt, turned that into a cascade plus anchor prompt. So that's what I'm doing here. Sorry if this was a bit messy. And this is the result. Okay, Scene two the reveal, it gave me the cascade with timeline, and I know I need to paste my two anchors plus the new corrected prompts with the tags. Okay? That's important. I have my tags and my anchors. So let me show you how this went in Canva AI. So this is scene two. Give me this video right here. Let's play. Then I pulled up with this bad boy. Max printing, baby. Okay, so here we start to see different problems. First, there is the hard coded captions right here. And then we have quite a different Gutenberg. Okay, so this was the first one, see, right here. And now we have this one. It also has a very different voice from the first one. So different hat, different outfit. So I was not super happy about this plus, also, there's some weird thing going on with his mouth at the end, you see this stuff there. So this is weird, and I didn't want to use this video right here. So what I did and what I suggest you do, I know it's annoying when you only have five credits to kind of waste a credit like so. But what I did is simply try again with the same rum. I re pasted the entire prompt, the same thing. I didn't change a thing, and it gave me this video. Then I pulled up with this bad boy. Mass printing, baby. So this was better. Still have the hard coded captions, which was annoying and also had little problem with the mouse going weird at some point, but that was better. Still quite different from my first character. And I didn't change the anchor. So I tried again, I think. Let me see. No. Actually this is where I decided, you know what? Kind of heard that I can also generate videos with Vio three in Leonardo. So I went there and checked it out. 13. Generating videos with VEO3 in Leonardo: I can actually come to Leonardo, and it will bring you straight here to your video feed. So before getting here, okay, let me show you kind of like the normal interface of Leonardo. It looks like soap, okay? It's very cool. I really like the vibe of this app. Okay, so here on top, you will have different options. You can obviously generate images with Leonardo. That's what it's famous for, and it's actually a great tool for generating images. But you can also generate videos. So when you click here, you will land on this interface, and you can choose different models, but you can see that Vo three is now available right here. So make sure you select VO three as your model. From there on, you will need to use credits. So I'm not exactly sure when you start if you get some free credits with Leonardo, I believe you do. I'm not sure if it's enough to generate videos with VO. Me as an ambassador, I have a bunch of credits, so I might as well use them. And these credits are called tokens or fast tokens right here on Leonardo. The good thing is that you can top them up, from Leonardo. So that kind of answers the question. Can I buy more credits? Well, not in Canva, but you can in Leonardo, which belongs to Canva. It's been acquired by Canva. Are part of the Canva family. So what you see here is my feed, my VO three feed in Leonardo. Who's this nice guy? Well, this is our good old friend Guttenberg, okay? Let's start with this one right here. So this is the one I ended up using in my video, as you might have noticed by now. And I'm going to click on that little arrow right here to reuse my prompt. This will load my prompt right here in the prompt box, okay? So I have my anchor. I have my ambience with the tags, and I have my cascade prompt structure right here. All the details. So I generated this video, cost me 2,500 past tokens. I'm not sure how much money that represents, but probably not cheap. And I got this video, which was great. And so I continued, right? What I was not able to do or what I was able to do, but didn't give me the best result in Canvas AI. I was able to ter right here in Leonardo. And the reason is, I believe, because Canva uses a version of VO that is lighter and less expensive than the model we are actually using right here in Leonardo. Okay? And when you do use Vo three in the Gemini app, you have these distinctions between the two models, like the two level of, I would say, horsepower that you have for each of the model. You have your VO three fast and VO three full, if I am correct. I have never used the Gemini app because I haven't paid for the 250 bucks per month, but that is what my friend Tianyu told me. So I believe what's going on here. According to TianU as well, he confirmed that to me, Leonardo is using the full model, while Canva is using the fast model. That might explain the difference in quality and in prompt adherence. So this was my second video. So basically, from this one, this one, and this one, it was like three out of three. Like no fails, every single prompt. So these prompts, I just continue prompting here, and I just took this, copied the entire prompt structure. Let me show you, like, the scene number three, for example. So I copied my entire anchors right there, and then the prompt for scene number three. Generate that it was a bunch of credit that kind of burned, but every single video was perfect. So, yeah, that's exactly what I did here. I downloaded these videos and brought them back to Canva, and this is what you see right here on this timeline. So once I had everything laid out on my timeline, I chopped them up kind of, like, sentence by sentence. I trimmed a few of the clips to make the transitions a little bit nicer, but not much, really. Like, not much editing at all. And Canva was more than enough to edit this little clip right here. 14. Generating captions automatically in Canva: So when I had everything kind of cleaned up, I used the text button right here to generate captions or auto generate captions. So Canva analyzed the video. Let me click here to show you. You see? You can select which video on your timeline you want to generate your captions. Me, I made sure all the pages were selected, and then I generated the captions, and Canva generated these captions that you see there. I style them. They don't come in that color by default, and they don't come at that exact same place either. So what I did, I just pulled out a guide from my ruler, and I placed it exactly where I position my captions. And I positioned them right on top of some hard coded captions that I couldn't get rid of. Also, what I did, let me show you here. For example, when you see this is scene one, which is not zoomed in. The scene one shot looks like soap. But what I did here, I actually zoomed in here by creating a cut right here, splitting my clip by using the S key on my keyboard to split the clip. 15. Tips and tricks to get rid of hard coded captions in VEO3 videos: Now, is there a definitive prompt hack workaround that will help you like, totally get rid of these hard coded captions? The answer is no. Okay? Every time you generate a video is kind of like throwing the dice, and you don't know exactly what's going to come up, okay? That is just the state of what the AI generating videos is as of today. We cannot guarantee there will not be these hard coded captions. These are inherent to the model right now. I'm sure Google is working on this. They are definitely aware of this, and it's definitely annoying a lot of people. You see here, my captions were hard coded. They are always full of typos. Sometimes they don't make sense at all. But one thing I have read again and again on different places, different videos I've watched, different conversations I've had with people in that niche, like the AI video niche. They told me again and again that one way of trying to lim probability of having these captions is to remove any quotation marks from your dialogues. Naturally, we tend to kind of insert the dialogue between error marks or quotation marks. That's kind of how we've been writing dialogues in scripts and books and stuff like that. But here, apparently, it doesn't help. So get rid of these quotation marks in your dialogue line, in your cascade prompt. This will help. It will not be a definitive hack. You will still sometimes have your hard coded captions, but I have also found that it helps reducing the probability. 16. Wrapping up and Canva Pro Free Trial: Alright, guys, this is what I had for you today. I think we've covered a lot of ground. This is kind of where I am. I showed you how to use the anchors to improve character consistency, how to pair these anchors with your cascade prom structure and to kind of use them together for scene and character consistency from scene to scene. We've discovered that we can use video in a different tool still from the Canva family, Leonardo AI. And we have attempted to get rid of these pesky closed captions that are showing in our videos. Or at least I give you a few tips on how to handle them once they show up in I've also showed you how to use auto generated captions to create some new captions based on your final video. That is a pro feature as well, but if you are generating videos in Canva, you must have a paid account. And if you don't yet have a paid account, but would like to try all of these cool features and also generate videos, scan this QR code. It will give you 30 days of free trial to Canva P so that you two can have fun with this tool. It's really fun. If you are not convinced you can cancel before the end of your 30 days, you won't be charged thing. But, yeah, I think Canva Pro is really, really, really becoming awesome in terms of the value it packs. So yeah, consider the investment if you're serious with your design or creator career. Alright, folks, that's it for me. Show me some love in the class conversation. Like, let me know what did you learn? What did you learn about VO three? What are the prompts that work for you? I'm really curious. Like, I've done this deep dive myself, but I'm here to share with you and with the rest of the community. Check out my LinkedIn profile. This is where I share all of the video created with View three. You'll see there are some fun ones, and don't forget that there is a class project that you need to complete if you want to graduate from this class. So make sure you go back to the class and read that project description. Have fun with it. You'll see this is a really fun one. Once you're done, you can create a public link or public view link in Canva to your video and share that in the class project description so we can all see your View three video.