Produktvideos nachbauen mit GPT-6 Astra, DaVinci Resolve und Blender

Hey guys. Now, generating videos and motion graphics with AI is now better than it's ever been before. People are literally grabbing real product launch videos and using Astra to recreate these videos one-to-one. And not only that, Astra can actually bring in that video, recreate it inside a video editing tool like DaVinci Resolve, and allow you to customize every single frame. It could also create 3D scenes in Blender and render videos based on those scenes. So in today's video, I'm going to be diving into these different topics. And if you like these types of workflows or if you like to learn more about these types of workflows, feel free to join my free Discord community down below. We get together every single week to do different types of co-working sessions and challenges. And by the way, I know that over 90% of you watching are not subscribed to this channel. So don't forget to subscribe if you want to keep updated with all of these new workflows. Anyways, the first thing that you want to do here is you want to create a new folder inside of Codex. You can use the ChatGPT desktop app. I have all of my GPT-6 projects in this folder that I have stored locally on my computer. And I can just open up a new chat that looks like this, and it's pretty much as simple as that. Now, the first thing that we want to do is we want to go from video to video. So the first thing that I can do is I can find my reference video, and let's say that we want to get this video. And it says "GPT-6 Astra can't match this quality of motion." And we can go ahead and download this video. And we can write something like, "Recreate this video with Hyperframes." And for those of you that don't know what a Hyperframe is, well, it says here it's an open-source framework that turns HTML into video. And I actually have a great tutorial on this, but I think it's a little bit outdated now because of GPT-6. But what you want to do is you want to go to your plugins and add this as a plugin. And that's it. Then we can go ahead and submit this prompt, and in just one prompt we get our video, right? It's not perfect. As you can see, there's like a little bit of a wiggling issue going on here in the middle, but overall, it's not bad, honestly. The text transitions look amazing and some of these movements actually look very nice. We even have the text being written out over here, and honestly like a lot of little sections that we can eventually duplicate or remix for another product video. And to make changes, all we have to do is just write it in the prompt. So instead of having green colors, we can make it more blue-toned. We can write something like, "Convert the green colors into a blue color tone." And then we just get the same exact video but now with a more blue tone instead of a green, and everything has been replaced. But for me, the craziest part about this experiment was recreating it with DaVinci Resolve because the quality of the motion graphics was much better. And you can literally just give it like a simple prompt like this, right? Attach the video again. "Try recreating the first 10 seconds of this video with DaVinci Resolve." And once I did that, my DaVinci Resolve app opened, and I also got this little guy over here summoned on the bottom right of my monitor. And so as you can see, I scrolled down, it took around like 20 minutes. And I go to DaVinci Resolve, and look at the animation of these little lines, right? Before with Hyperframes, they were a little bit more static. They wouldn't curve like this, but now that curvature is actually happening and it looks very smooth. And by the way, I personally do not know a lot about DaVinci Resolve and I use this tool to not only help me learn new software, but also help me prompt, which I'm going to show you in a minute. I use this tool called HeyClicky and they have a really, really good free plan. So everything that you see in this video is going to be done with that free plan. And basically what it is, it's this little blue cursor that's following my cursor. And I can hold down control and option at the same time and then I can just talk to Clicky. So, "Hey Clicky, I'm here in DaVinci Resolve. I want to edit the text in the composition. I just don't know where to do that." So select the composition clip, then click Inspector. It's editable text controls should open on the right. Click the fusion page so we can access the text nodes inside this composition. Click this text node in the upper left node chain. Click modifiers to reveal the editable text control for this node. And so here, I can basically edit the text, but what I like to use Clicky for is actually prompting. So I can find a specific frame of this video and I can then hold control and option again. "Hey Clicky, so look at this frame. I need you to help me write a prompt so that you can add it to my clipboard and I can just paste it into Codex. I need to basically tweak this frame. Right, right now it says, 'You don't think about finding your way home.' I want it to say, 'You always think about looking at art.' And I want the background to be black, I want the text to be white. I want these little stripes to be red, yellow, and black. I want the text to be 10% larger and I want it to be a serif font. And, yeah, just add that all to my clipboard as a prompt." And so I can go back into Codex and paste in that prompt and also screenshot that frame. And so you obviously want to read through the prompt. I read through it, it looks good. And we can send this. And as you can see now our video has now changed. Black backgrounds, we have these new yellow, orange types of strings coming up with the new text. You know, just an example of what you can do. But another example of a video that I wanted to clone was this one released by OpenAI to announce GPT-4 image to images 2.5. It's a super nice video, obviously, by a huge company. But I really wanted to see how Astra could copy a video like this. And so again, I went into Codex and just said "Recreate this video." I added that file as a reference. And so again, almost one-to-one, really. I think there are just like a few little things that Hyperframes didn't do perfectly. But I imagine that if you were to give this to DaVinci Resolve, it would do a better job. And for example, we have like sketch L. I don't know why it added that L. But I was like, "Hey, what if we change a few things from this ad? Like maybe we can replace this drawing with a random drawing that I'll make." And you know, this is about the sketch feature, so you can sketch something out and then like different images will appear. And so I wanted to sketch like a random little character and have have that video create three different images with that character. And so this is the little character that I made. I imagine him as like a little orange fruit and his eyes, his mouth, his arms, his legs, and this little his little hair is made out of clay. And so instead of me having to write all of that, I can again just summon Clicky by holding down control and option at the same time. "Hey Clicky, so I'm trying to write a prompt for Codex to basically replace the first character in that video that I just generated with this sketch over here that I'm going to add as a reference, and I want to create three different scenarios. One scene with a neutral background, one scene where this orange dude is in a laundry store, and another scene where he's in a pool. Keep the style is very consistent. Important to tell Codex to also keep the our orange guy consistent. And just write this as a prompt or just write, or add this to my clipboard so I can paste it as a prompt into Codex." So it's analyzing the image. I wrote the full Codex prompt with the three scenes and strong character consistency constraints. And so I added the screenshot of our little character. I pasted in the prompt that Clicky gave us, and you just want to obviously read through everything. But basically, it says to use this as a the visual reference for the main character, to replace the sketch with this one, and then add these three new scenes. And to also keep the character's appearance extremely consistent and to match the visual style. And so then we can go ahead and click on submit. And by the way, some of you might also be wondering why I don't use voice mode from Codex. Problem is that in voice mode, it's more of a conversation. So you submit a message, then you get one in return, then you submit another one, and so on and so forth. So you're wasting a lot of credits with that conversation. And with Clicky, you can just rant and rant and pause and rant and at the end, everything that you said turns into one message that you can just send to Codex. But anyways guys, this is what we get. So "sketch," we have our little sketch. It looks identical to what I sketched out in Excalidraw. And now we have our little orange with the neutral background. We have him in the laundry store, and we have him in the pool. Look at that. All with a very consistent style. And you can also edit these types of text with this like little handwriting, handwritten text at the bottom. So for example, in this scene it says "create artwork." This is from the original video. And you can just change the text to be like "create modern art," for example. And now another cool trick that you can use with Codex is actually giving Codex its own infinite canvas. So for example, you can just post that video that you just created into this canvas and you can ask Codex to like render every one second as an image and just place it over here, you know, on the side so that as you talk to Clicky, you can like write different notes like, you know, "change this, make this pop out like a little bit more, like so, change this one to blue, make this a swamp instead of a pool." I mean, you can say whatever you want, you can write whatever you want, but you know this is a plugin called Magic Path and I've mentioned this tool a bunch of times in my channel. And at the end of this video, we're actually going to use that method to generate a video, a remix video from different videos that we kind of copy from the internet and make a fictional product launch video for a another random brand. But before we get to that, another plugin that we can use is the Mobbin plugin, and I can tag Mobbin by writing @mobbin. I can write something like, "Find me a UI card for a fintech company that has 3D icons." And basically, Mobbin returns different results to me, and one of the results was this one, which is actually perfect, so I want to I want to basically get this little card and animate it somehow. And now for this example, I'm going to be using a tool called Diffusion Studio. They also have a great free plan, by the way. And so what I can do is I can just copy-paste the screenshot of that image into Magic Path, select the image, and I just can write, "I need you to recreate this UI in Diffusion Studio." Because this Magic Path plugin can detect what you're selecting in any specific file. And I can also in Magic Path, draw out like scene 1, scene 2 of what I want this animation to look like for this UI card. And so I can then, control and option. "Hey, Clicky, so this is a UI card that I want to basically animate and these are the different, this is the animation down here, right? This is scene one and this is scene two. So I want it, I want this card to remain the same size and I want it to start off like in a position down at the bottom right and then, nice a nice ease-in, ease-out movement with like some motion blur, pop up to the center of this viewport. So, can you please create a prompt for me? This should be around like 3 to 5 seconds long and add that prompt to my clipboard so I can just paste it into Codex." And so we have our Diffusion Studio UI like this. The video is about 5 seconds long and we have that exact same animation that we wanted from before. You know, just a quick simple example of what you can do. It has that motion blur, it has that nice ease-in, ease-out movement and most importantly, it's the exact same card that we got from Mobbin. And everything here is modifiable. Like we can edit the text, we can edit the icons, and also I can ask the Codex to build this card out in Blender and create like nice 3D animations in Blender. And so I found this like perspective screen in a video on YouTube. It has like this nice blur over here and over here. And I just literally took a screenshot of this like so, and I can then add that to the prompt and also ask Clicky to describe what it's seeing, and it creates the actual Blender model. So as you can see, this is like a real 3D model that you're seeing right in front of you. They even have the numbers over here, they have the logo, they have like the colors, the the the texture. It looks very precise, and this is the final video. And so now that we've covered the Diffusion Studio, Hyperframes, DaVinci Resolve, and Blender, I wanted to do like a quick final little project here combining everything. And so I found this product on Amazon. It's a keyboard, it's this brand called AULA, and what's crazy is that you can just say, "Use Blender to turn this into a 3D model," and you just paste in that link from Amazon, and this is the model that you get. I mean, you get the full product. They even have the numbers over here, they have the logo, they have like the colors, the the the texture, it looks very precise. And so I asked it to make like a little 5-second animation where the keys are separated and then and then they slowly come back to like the base. And to be honest, it's not a very special thing. Like, you know, you'd expect GPT-6 to do this shockingly. But going back into Magic Path, I kind of got some screenshots from that GPT-6 video that I showed you all earlier, the 3D model, the logo of the keyboard, and just some like ideas of what we can do for each scene. And so first, you want to like, click on this dropdown button, then this pops up, then it selects this blue color, and then there's a top perspective of this view where the keys are blue and then we have this scene over here, everything fully created in Blender by the way. And then at the end, the logo that kind of slightly fades in and boom, it's 10 seconds long. So anyways, guys, hope you enjoyed today's video. If you have any comments or questions, please let me know down below. And again, feel free to like and subscribe if you like this type of content. I go live every single week to kind of go through these different workflows. So if you want to join that, link to that is down below. But anyways, thank you all so much for watching. Hope to see you next time. Goodbye.