07:39Zane Hoyer · Tutorial
AI Motion Graphics Just Got Scary Good (Full Workflow)
A 7-minute walkthrough of five motion graphics use cases built entirely with Claude Opus 4.7 and Seedance 2.0 — no After Effects, no keyframes.
May 26thThe argument in one line.
Directing an AI agent that accumulates project context and selecting the right image model per shot, then finishing the output with three Resolve effects, produces a broadcast-quality intro in roughly twenty minutes of active work.
Read if. Skip if.
READ IF YOU ARE…
- You edit YouTube videos in DaVinci Resolve and want to stop spending hours sourcing royalty-free B-roll.
- You have tried text-to-image or text-to-video generators but found the outputs visually inconsistent across a project.
- You are comfortable in Fusion's node editor and want step-by-step animation recipes you can apply directly.
- You make documentary-style or educational YouTube content and want a high-budget-looking intro without a full production budget.
SKIP IF…
- You work in Premiere Pro or Final Cut — the entire Fusion section is DaVinci-specific and does not transfer.
- You want a purely manual or AI-free production workflow.
- You are brand new to DaVinci Resolve — the Fusion walkthroughs assume familiarity with the node graph.
The full version, fast.
The tutorial presents a two-stage cinematic production engine. Stage one uses InVideo Agent One: you describe your project, paste your script, attach reference images, and the agent builds a shot map, selects the appropriate image model per shot (GPT Image 2 for text, Nano Banana Pro for photorealism, Recraft for skin), generates images and short video clips, and runs a final-cut review loop. Stage two moves into DaVinci Resolve: two Fusion animations (a folder widget and a notification widget built with bouncy double-axis keyframes) replace the weakest AI shots, and three stacked adjustment-clip effects (edge detect glow, a dot-matrix scanline overlay, and film grain with zoom blur) elevate the remaining AI footage to a premium cinematic look. Total active time: roughly twenty to thirty minutes.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:22
01 · Intro and showcase
Documentary crime-intro plays as proof of quality; thesis stated: AI agents plus editing skills beat AI-only output and AI-slop channels.
01:22 – 07:57
02 · Agent One workflow
Full InVideo Agent One demo: project setup, script paste, shot map generation, reference handling, model auto-selection, video generation, task-tracker overview, final-cut feedback loop, context tab accumulation.
07:57 – 14:39
03 · DaVinci Animation 1 — Folder widget
Fusion composition clip: iOS wallpaper background, vignette, motion graphic overlay, app-bar screenshot, quick bevel, bouncy text, animated mouse cursor with curved spline path; adjustment clips add edge-detect glow, dot-matrix scanlines, dust overlay, camera zoom ease.
14:39 – 18:50
04 · DaVinci Animation 2 — Notification widget
Grid background, rectangle mask with double-axis asynchronous bounce keyframes, write-on text animation, Reddit logo, mouse icon; duplicate-and-drag workflow for multiple widget instances.
18:50 – 22:50
05 · DaVinci Cinematic Effects — AI footage finishing
IntelliTrack four-point tracking for match-move colored card overlays with outlines; saturation to zero with colored shapes; film grain, zoom blur, and film damage on adjustment clips.
Lines worth screenshotting.
- An AI agent that accumulates project context gets faster and more consistent on every subsequent video in the same series.
- The agent automatically picks GPT Image 2 for shots with on-screen text and Nano Banana Pro for photorealistic shots, so you never choose the model manually.
- Feeding a finished video back into the agent as a review object and asking what is not working closes the quality gap without manual note-taking.
- Offsetting width and height bounce keyframes by two frames makes a widget animation feel organic instead of mechanical.
- Edge Detect in ResolveFX at low brightness adds a dreamlike glow to AI footage without making it look post-processed.
- Two perpendicular scanline layers at high frequency and low sharpness produce a pixel dot-matrix look that reads as a real computer screen.
- Dust overlay at 8 percent opacity on a Screen composite mode adds film texture that is invisible at normal viewing distance but registers subliminally as cinematic.
- IntelliTrack on four corner points gives you a match-move anchor that keeps colored card overlays locked to camera movement in AI-generated video.
- Turning saturation to zero on AI footage except for colored tracking shapes makes the shape annotations pop against a black-and-white background.
- The compounding value of an agentic workflow is not the first video but the fact that the agent trained context makes the second video faster than the first.
How to direct an AI agent like a producer, not a prompter.
WHAT TO LEARN
The gap between AI slop and cinematic output is not better prompting but giving an agent a persistent world to work inside and then finishing the result with two or three specific Resolve effects.
- An AI agent that stores project context (visual language, rules, knowledge bank) produces more consistent outputs than one-off prompting because it references its own accumulated decisions.
- Reference handling, uploading real images and asking the agent to extract lighting and framing separately, produces shots that fit your visual style instead of averaging all references together.
- Feeding finished video back into the agent for a written critique and then confirming the changes closes quality gaps faster than iterating on prompts alone.
- The folder widget and notification widget animations share the same underlying Fusion pattern: rectangle mask with bouncy double-axis keyframes, quick bevel for depth, and write-on text. Learn it once and apply it to any UI animation.
- Offsetting width and height keyframe timing by two frames is the single change that makes a widget bounce feel organic rather than mechanical.
- Edge Detect at low brightness, two perpendicular scanline layers faded out, and a dust overlay at 8 percent opacity are three low-effort Resolve effects that collectively make screen-content animations read as cinematic rather than tutorial-style.
- IntelliTrack with four corner points gives you a match-move anchor in AI-generated footage so colored card overlays stay locked to camera movement rather than drifting.
Terms worth knowing.
- Agent One
- InVideo's agentic product that maintains a persistent world (context, visual language, knowledge bank) for a video project and dispatches sub-tasks autonomously once given a direction.
- Shot map
- A table generated by Agent One that breaks a script into individual sentences and assigns a visual concept and generation prompt to each, acting as a pre-production storyboard.
- Reference handling
- A feature in Agent One where you upload real-world reference images and instruct the agent to extract specific lighting and framing elements rather than averaging all references together.
- Fusion composition clip
- A DaVinci Resolve clip type that opens the Fusion node-based compositor, allowing complex layered animations to be built entirely inside the NLE without leaving the timeline.
- Quick Bevel node
- A third-party DaVinci Resolve Fusion node that adds a subtle lit bevel and gradient to a shape, simulating depth without a separate 3D render.
- IntelliTrack
- DaVinci Resolve's AI-assisted motion tracker inside Fusion that automatically follows a user-placed tracking point through a clip, enabling match-move overlays on shaky or zooming footage.
- Match move
- A compositing operation that locks an overlay element to tracked camera or object motion so it appears physically attached to the scene rather than floating on top.
- Edge Detect (ResolveFX)
- A DaVinci Resolve effect that isolates high-contrast edges and creates a glow around them, used here at low intensity to give AI footage a dreamlike luminance.
- Film damage (ResolveFX)
- A DaVinci Resolve effect that simulates aging film artifacts such as moving scratches, dust specks, and vertical lines to make digital footage read as archival or degraded.
Things they pointed at.
Lines you could clip.
00:42
“You still need editing skills to keep that high budget feeling and to separate yourself from the AI slob channels.”
07:10
“We are essentially in the process of training an employee specific to our needs.”
06:50
“In the span of, like, twenty minutes going back and forth with agents, we've created something that would take hours to do on your own.”
14:04
“That took us, like, five or ten minutes. It was not complicated at all, and it looks better than 99 percent of the stuff on YouTube.”
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
00:00The one goal I had for this video was to make a documentary intro with better editing than anything else on YouTube. This tutorial combines both in-depth DaVinci Resolve animations and a high end AI animation workflow. Add in good music and sound design, well, you tell me if you'd watch this video.
00:16What actually makes a criminal a criminal? Is it a fixation on weapons, the environment, or a suppressed childhood memory?
00:28I've spent the last month tearing through forgotten crime scenes, analyzing hours of unreleased interviews, and tracing footprints that are supposed to be buried because the truth isn't in the obvious evidence. It's in the connections everyone else refused to see. Let's look under the hood.
00:42As AI gets better, you start to realize adding it into your workflow is mandatory. But you still need editing skills to keep that high budget feeling and to separate yourself from the AI slob channels. 70% of you guys aren't subscribed.
00:55So if you find any sort of value in this video, thank me by hitting that subscribe button and the like button. So not only are AI video and image models improving every month, there's also AI agents that are drastically improving. And an AI agent can be trained on a specific topic.
01:09But essentially, it can work and think on its own once you give it a task. It gets rid of the need for prompting constantly, regenerating, and repeating the same tasks.
01:17It basically gets rid of the hard boring part of using AI, or I guess what you would call the labor. So I'm using an extremely powerful model inside Nvidia called agent one.
01:28A link to it is in the description. So as you can see, one of the options here is to make an explainer video, and that is what got me interested. So I basically just described the time link, thirty or forty seconds.
01:37I set a crime documentary intro. Here, I'm kind of describing the narrator. I'm saying that he's kind of cold hearted, wedding photos, criminal investigation, interviews, forensic lab, weapon photos, etcetera.
01:47And then I pasted in the script of what I actually want him to say. And then I also said, see my attached video for the style I'm going for. And now we can just click this arrow.
01:55It wants us to name the project, so I'm just gonna put criminal investigation intro. So here is our first agent that we're basically building, so we can name it. I'll just name it agent one.
02:03So here it's created this chat for us, and it's gonna process the video file and the script. And so this tab over here that says notebook is where it's gonna process all the photos and videos. And then this tab over here under context is kind of the world that we're building.
02:17So this is gonna fill out over time based on how we collaborate with our agent. Obviously, there's not gonna be anything here yet, but this is like the general overview of the project, the rules, the agents, the world building inspiration, visual language and brand guidelines, and then knowledge bank.
02:31So you can start to see how in-depth this takes things. So it's gone over everything now, and here's what it's given us. It's gonna be thirty or forty seconds, 16 by nine, and ten eighty p, exactly what we want.
02:41Here is the overall vibe that it understands, and then here it's generated a shot map. So it's basically broken our script into different sentences, and then it's given each sentence a visual that it's going to generate. And then now it wants two things from us before it starts.
02:53The voice over, we can have it generate a new voice over or use the one in the video we provided, and then it's asking if we want on screen text. So now it's gonna start generating prompts for all of those shots.
03:03Now keep in mind, we're not gonna be using all of these shots because we still are making our own animations. And when you combine the two, it creates something that is extremely high quality. Here, it's starting to generate some images and it also gave me three choices of voices.
03:17And so once I pick the one I like, I can select finalize. So as you can see, it's starting to build out the images for the shots and these are exactly what I had in mind.
03:27So these are good. And then it's also generated the voice over here. Something I'm gonna do for the actual interviewing shots is use something called reference handling.
03:37So I can upload three screenshots I've collected of criminal interviews I just got from Google. And so what I did is first, I confirmed that that these photos are good.
03:47I said that the criminal evidence photos are perfect. This is the exact realistic vibe I'm going for. Now as for the criminal interviews, I've uploaded three screenshots.
03:56I want you to pull specific elements from each, like the lighting and the framing, and I want you to use your best judgment to create a very realistic but eerie interview shot. Now, it's not just gonna look at these three images and mush them together. It's actually going to tie our entire vibe into these images, and essentially just design what it thinks will look best, and then we can either confirm it or brainstorm further.
04:15So my nose is actually bleeding, so I'm just gonna let my agent do the work. So as you can see, it's generated my interview photo.
04:24That is looking really nice. And then it continued down the shot map here generating the other photos. So like things like wedding photos, suspect photos, more forensic photos, and then a few more photos for the in between shots.
04:36And you don't have to worry about selecting the model. It'll decide which model to use based on your prompt. But as a general rule, if your image has text in it like this, it'll use GPT image two because it's really good at not messing up anything with text and getting really clear looking text.
04:51Another example of GPT image two is this, like, it's never gonna mess up any text just because it's so good at that. And if it doesn't have text, it's gonna use Nano Banana Pro a lot of the time because it's good at piecing everything together.
05:01And in really specific cases, it might even use Recraft for skin texture. So it really just depends on your prompt, and then it'll do the same thing for video. It'll select which video model it thinks is the best.
05:11So now that we have a lot of the storyboard built out, I can ask agent one to turn these into videos, and then once they're in videos, we can connect them together. So what I said is using the shot map, let's start turning these images into videos. Start with the forensic evidence.
05:25I want a fast paced video cycling through different pictures of evidence. Each frame should last about half a second. And then the same thing with the wedding photos, I want it to look like cycling through the different wedding photos.
05:35I want the criminal investigation scene to look like it's a real time lapse, and then I can hit send. And so because c dense two point o is so advanced, this is gonna look like a real video editor did it. It's not gonna look like some weird AI slop.
05:46It's gonna look as if a video editor clipped together a bunch of photos, which is gonna save you a bunch of time. So to give you an idea of how much time this is saving rather than just using a normal image or video generator is what an agent does is create tasks. As you can see, it says 21 out of 23 tasks done, and you can scroll through them, and it will mark them done as it completes them.
06:05So when I asked it to create these videos, it realized it needed more frames in the similar style, so it created those. And then it created the music and sound effects, and then it built the video.
06:15Same thing with the wedding. It realized it needed more wedding pictures, so it generated those. And then as you can see, it created the video for us with the music and sound effects.
06:22Now once it gives us these videos, one of the last things I do is the final cut review is I'll download the video, and then I'll reupload the video back into here. And I'll basically ask it to analyze the video here, and I'll ask it to tell me what's working and what's not, and I'll remind it that this is a fast paced YouTube intro.
06:38And then when I hit send, it'll now analyze that video, and it will come back with feedback. So as you can see, it's analyzed the video, and it's come back with feedback. She tells us what's working, what she'd fix, and then what she's actually gonna add to the videos.
06:51Now to confirm it, I can just say something like, great. Please make these changes. And then now agent one will begin fixing that video, and from there we can download it.
06:58So in the span of, like, twenty minutes going back and forth with agents, we've created something that would take hours to do on your own. You have to start thinking of all the research you'd have to do and all the photos you'd have to grab and make sure they're royalty free. At least in my case, when I'm editing videos and trying to find b roll, it's always just better to use AI now.
07:14One, it's quicker, and two, it just always looks better. Also, if we check back in our context tab here, you can start to see it fill out. We are essentially in the process of training an employee specific to our needs.
07:25You can see how collaborating with an agentic workflow saves even more time than back in the day when you had to do all the prompting and guesswork yourself. Plus, you get better results because whether you realize it or not, we just trained an AI agent on the exact style that we want. And now when it comes to making the next video, the process is even faster than before.
07:42This is the first time I've used it, so obviously, it has the potential to go way more in-depth, which I will definitely do in the future. So once you continue to do this entire process for the shot map here, we can now hop into DaVinci Resolve, pick the parts that we don't love, and replace them with some beautiful animations.
07:58So even though these animations look amazing, honestly better than most videos you'll find on YouTube, these animations are not that complex or hard to make at all. So I'm gonna walk you through how I made them, and then I'm gonna show you some effects I put over top of them as well as the AI footage we generated to give everything a really premium cinematic look.
08:14So we're gonna start with this folder animation because it pretty much works with whatever video you're making. So gonna come to effects and drag on a fusion composition clip. Now we can go into the fusion tab, and we can drag in a background and connect it like this.
08:27So I'm gonna drag in this screenshot of a wallpaper. It's just like a iOS wallpaper. By the way, all the assets I'm using are in the description.
08:34I'm gonna hit shift space and add in a transform, and then we can just size this up. Now I'm gonna hit shift space and search v I, and I'm gonna select vignette, and then I'm gonna add that in.
08:44And then holding down shift, I'm gonna bring this into the timeline. Now we can drag up the softness just like that and increase the size just a little bit. The idea I have behind this is the viewer will be more focused to whatever we put in the center of the screen.
08:56Now I'm also gonna drag in this motion motion graphic. This is also in the asset pack. Now if we connect this to the merge here, and now we can hit shift space and add in a transform, and we can size this all the way up like that.
09:09Now if I click on it and hit shift space and search luma here and add that in, it's gonna get rid of all of the black. Now it's still a little bit too visible, so I'm gonna drag in a background like this and connect it to that media in.
09:21Now the more we drag down the alpha, the lighter it becomes. Now something I did to save time is I took a screenshot of my app bar down here, and then I'm gonna drag that in here and connect it to this merge. Now obviously, we have some rough edges here and it doesn't look great.
09:37So what I'm gonna do is select the rectangle mask, and then we can drag up the width and drag up the height. Then we can drag up the corner radius. So now as you can see, it looks a lot more clean.
09:48And then to add a bit of a bevel back to it, we can hit shift space and search quick bevel. If you want access to this node along with a bunch of bouncy text animations, the link is in my description.
09:58So I'm going to drag the spread all the way down and then I'm going to drag the light power down as well. And now as you can see, we have a nice bevel.
10:06So I'm gonna hit shift space, add in a transform, and we can just drag the size up and place it down here. From the asset pack, we can drag in this folder icon and we're gonna connect it to the vignette right here. Gonna hit shift space, add in a transform, and size it down.
10:21Now again, I'm gonna hit shift space and add in a quick bevel. And we can just zoom in here, turn the spread down. I'm also gonna select the transform, hit shift space, and add in a drop shadow.
10:31I'm gonna drag these two up like this because we want the text underneath here to be above the transform. So I'm gonna select this, hit shift space, add in a merge node. And now I'm gonna hit shift space and add in a super bouncy I'm gonna add that in, and I'm gonna connect it to this merge.
10:47Now if we play that back, you can see the text is already animated like that. These bouncy text animations are in the same link as this quick bevel. I'm gonna hit shift space, add in a transform, and size this down, and bring it right here.
10:59And we can label this whatever we want. Now I'm gonna drag out my media out and because this one's in the middle and bring it in the middle. And I'm gonna select everything, hit command, copy, paste it, holding down shift to bring it into the timeline, and then reconnect our quick bevel.
11:13Now we can select the transform and just drag everything over. I'm gonna select the text and change this to number one. And then again, I'm gonna select this, command copy, look over here and paste it.
11:23Holding down shift, we can bring this into the timeline and drag it over. Now in our asset pack, I'm gonna drag in this mouse cursor icon and connect it to this merge. Hit shift space, add in a transform, size this down, come to the first frame, and we can just bring this off green for now like here.
11:39Make sure we keyframe the center x and y position. And then we can come over to, I don't know, whatever you want, maybe frame 85. And we can just drag our mouse to where we want it to end.
11:48We wanna go to settings and turn on motion blur. I like to select the transform, select the edges of the starting point, and then you can bend these like this. And select this one and bend it like this.
11:58And then that way the path of our mouse is curved instead of just like a straight line. You also wanna make sure you go to the spline, select a box over everything, and press s. And now it's gonna start slow, speed up, and then end slow.
12:09If you want these to be different colors, select one of the media ins, hit shift space, and add in a color corrector. And now you can just change the hue to whatever we want. Now as you can see, we've taken the concept of someone's computer screen and made it kind of cinematic.
12:23So to make it even more cinematic and make it look a little more polished is we can come to the edit tab, come to effects, and drag on an adjustment clip. First thing I'm gonna do is go to the color tab and just increase the saturation. And I'm gonna come to effects, resolve effects.
12:36I'm gonna search edge and select edge detect and drag it on the adjustment clip. I'm gonna select this box, edge mask overlay, and I'm gonna increase the blur and the edge width.
12:47Now, when you play around with the brightness, you don't want it to be too strong, so I'm gonna turn it down quite a bit. Everything is glowing a little bit more, so you can turn this on and off to see the before and after.
12:57But this just makes it a little bit more dreamy. Now another thing I'm gonna do is drag this up and I'm gonna drag another adjustment clip right here.
13:06I'm gonna create some scan lines to make it look like a computer screen, but the reason I'm putting it on a second clip is so that I can fade it out. I'm going to come to ResolveFX and search scan, and I'm gonna drag on scan lines.
13:17I'm gonna come to effects on our adjustment clip here, and I'm going to increase the line frequency a lot. And then I'm going to drag down the line sharpness just so it's not so sharp on our eyes. Then I'm going to drag that on again, but this time I'm going to rotate the angle 90 degrees, increase the line frequency, and decrease the line sharpness.
13:35Now if we zoom in, you can see we've created these little dots here and it looks like a real computer screen now. Gonna take this little white dot and drag it. So now it starts out with these dots and then fades out to nothing.
13:47I'm gonna drag these two up like this, and I'm gonna drag in this overlay of some dust. This is also in the asset pack. So I'm going to zoom in like this and I'm going to change the composite mode to screen.
13:59And obviously, this is too much dust, so I'm gonna turn the opacity down to like 8% so you can just barely see it. So if we zoom in, you can just barely see these dust particles, and it makes it look really cinematic.
14:10So the final step is we're going to animate a camera zoom with this top adjustment clip. So on the first frame, I'm going to keyframe the zoom, and then come to the last frame and zoom in. Now I'm gonna come to keyframes up here and select this up here.
14:26I'm gonna drag a line over the two keyframes, select this option, and select ease in and out. And that way, creates a nice smooth zoom. So as you can see, that took us, like, five or ten minutes.
14:36It was not complicated at all, and it looks better than 99% of the stuff on YouTube. But this animation is also really fun to make, and it's not that hard at all. So, again, we're gonna drag on a fusion composition clip into the fusion tab, drag on a background.
14:49We can select the color and make it slightly lighter. I'm gonna hit shift space and add in a grid. Select the grid, add that in, and I'm going to increase the row cells and the column cells.
14:59I'm gonna change the color to something really dark just so they're barely visible. And then I'm gonna hit shift space, search v I, add in a vignette. And again, we're gonna increase the softness.
15:09So there's kind of our base background. So now we can make that first pop up widget with the text. And once we make one of them, we can just duplicate it.
15:16So it's pretty simple. So we're gonna drag in a background connected to this here. We're going to select the rectangle mask and make the shape of our widget.
15:25So I'm gonna do something like this, increase the corner radius, something like that. Select the background.
15:30I'm gonna increase the lightness something like that. But to make it actually visible, we can hit shift space, add in a quick bevel.
15:39And so this first one, we're gonna actually gonna keep it like that, but we're just gonna turn down the light power. And so now as you can see, it looks like it has a bit of a gradient. I'm gonna hit shift space and add in another quick bevel.
15:48In this one, we're gonna turn the spread all the way down and then turn the light power down and we end up with a widget that looks pretty nice like that. And so before we do anything else, we're going to keyframe kind of a bouncy animation in both directions. We're gonna turn the width all the way down and key frame it.
16:04We're gonna go over 10 frames. One, two, three, four, five, six, seven, eight, nine, ten. Increase the width.
16:09We're gonna go over nine frames. One, two, three, four, five, six, seven, eight, nine. Drag it down a little bit.
16:14And then eight frames. One, two, three, four, five, six, seven, eight. Drag it up a bit, seven, one, two, three, four, five, six, seven, drag it down.
16:21We're going to come to settings, turn on motion blur. We're gonna come to the spline, select it, and press s to round everything out.
16:30Now when we play that back, it looks like that. But still a little bit boring, so we're gonna also animate it on the height side of things.
16:37So on the first frame here, under controls, we're going to also keyframe the height. We're gonna drag it all the way down and we're gonna go one or two frames past frame 10 because we we don't want it to be synced at the same time.
16:49So on frame 12, we're going to increase the height past to where we want it. We can go to the next key frame and two frames after it and drag the height down. And then again two frames past our other one, drag it up.
17:03We're basically creating the same animation just a frame or two ahead of the other one. So now if we watch that back, it kinda animates in two different angles and it looks really bouncy.
17:13So from there, can just drag these up and add in a merge node. So from here, we can add in like a text animation, connect it to the merge, write a name, and then one hour ago. And we can just change the color of these, add in a transform, and just place it where you want it.
17:30Add in a merge. This time we can just do a normal text node like this, and we can type in a sentence like add in a transform, size this down.
17:38Again, we can change the color to kind of a more gray, something like that. And so now we don't want this to write out until it's formed.
17:45So maybe frame eight, we can select the text, keyframe it, come over to how long we want the animation to last, so like frame 46, keyframe that area, and then we can hit this arrow to jump back to the first keyframe and just drag this all the way down. And now starting on frame eight, it'll write out till frame 46.
18:03And so now we can just simply select all this, hit command copy, click over here and paste it, hold down shift, bring it into our timeline, and we wanna reconnect our bevels. We can hit shift space, add in a transform underneath everything, and just drag it down.
18:18Drag our media out, select everything, command copy, and paste it in here, and we can drag this one up.
18:25And now again, you can add in the mouse icon that I add in a transform. And in the animation in the documentary, I also added this Reddit logo, add in a transform, and place this where you want it.
18:38And now if we go to the edit tab, we can select all three of these things we made last time, hold down option and drag it over. And just like that, everything now looks really cinematic, especially on a dark background that edge detect creates a nice glow over everything. So now we have all this AI footage we generated.
18:54I wanna show you how to make it look really good. So for example, here's one of the videos we generate. If we hover over our clip and select fusion, for example, I think I had each object selected by like a colored card.
19:05To do that, we actually need to track the camera movement first. We're gonna hit shift space and add in a tracker, and you're gonna see this little thing pop up. So I'm gonna select the corner of the frame, and then I'm going to select IntelliTrack to add another one.
19:17And I'm gonna drag it over this corner. I'm gonna select it again, put this one over this corner, and then one more time, this corner. And as you can see, I'm in the middle of the frame, so I'm just gonna select this middle one which tracks forward and then backward.
19:28We select operation and then select match move. So now whatever we connect to this tracker will essentially be tracked to the motion. For example, we can add in a background and connect to that tracker, Select the rectangle mask, and we can play around with the width and the height.
19:43I can change the color of this to kind of a dark red, and I can turn down the alpha on both of these so that we can see through it. And now if we watch that back, you can see that it's tracked to our footage.
19:53So if we go to the first frame, decrease the width, keyframe it, and we go to say, I don't know, frame 40, we can increase the height. Now we can go to settings and turn on motion blur, select the spline, select this, Select a box over this and press s, and I'm gonna drag this side up like this. So after you go to the first frame and watch that back, it looks like that.
20:13Now I want this red to kind of have a white outline. I'm gonna hit the background, hit shift space, add an emerge node, and I'm going to copy these two. Hit command copy.
20:21Click over here and paste them. And I'm going to connect this to the merge, and that basically just duplicated it. Now I'm gonna select the rectangle.
20:28I'm gonna select the solid, and I'm going to increase the border width. I'm gonna change the background color to white. Say you wanted to add two more, we can just select all this and drag it up because remember everything needs to be connected to this tracker.
20:40Select the merge, hit shift space, add an emerge node, select this background like this, paste it, and connect it to that merge. Now we can select the rectangle and using the center x and y position, we can bring it to this stuffy over here.
20:53And we can come to the last frame of our key frame and drag the width down. Again, I'm gonna hit shift space at an emerge node. Now we can copy this and paste it to this merge.
21:03Change the color to white. Select the rectangle and turn off solid, and increase the width. Now we can come to this background and change the color to kind of a green maybe.
21:12Now we can select the merge and hit shift space and add in a transform, and we can just rotate this angle and kinda place it like this. Now if we watch that back, these both get selected like that. And now you can just repeat the same for the sock.
21:26Hit command copy, paste it, add an emerge, connect this to here, select the transform, fit it to the sock.
21:34Change the background color to maybe yellow, turn down the alpha. Now say we drag in another AI photo we generated.
21:41I'm gonna quickly go into the fusion tab on this again, and I'm going to hit shift space and add in a color corrector. And I'm just gonna turn the saturation all the way down.
21:50So it's black and white except for the colored shapes. So I'm gonna take an adjustment clip and put it over top of this AI footage. And the first effect I'm going to add is called film grain.
22:01I'm gonna apply this to the adjustment clip. I'm going to increase the grain strength, increase the opacity, and then now we can see it very well.
22:10And we can play around with the texture, grain size. I'm gonna make it really small.
22:14It makes kind of a grain effect. It looks really good. Gonna come to effects under resolve effects.
22:18I'm going to drag on a zoom blur. Drag that on. I'm just gonna turn that down a little bit so it's not too much.
22:24It just subtly blurs the outside. If you want it to further look damaged, you can search film damage, drag this on, and it's gonna create lines and specs that move.
22:34And of course, you can adjust this further. And then the last step is sound design, And that's my next tutorial. I'm going to, uh, show you how I find all the best sound effects.
22:43So make sure you subscribe so you don't miss that tutorial.
The Hook
The bait, then the rug-pull.
A bold claim anchors the open: a documentary intro that out-edits anything else on the platform. What follows is a live proof-of-concept, twenty-three minutes of actual workflow from a cold prompt in an AI agent interface to a finished cinematic intro polished inside DaVinci Resolve.
Named ideas worth stealing.
07:10model
Agent Context Loop
Every interaction fills the Context tab: project overview, rules, agents, world building, visual language, brand guidelines, knowledge bank. The agent's trained state persists across sessions so each subsequent video starts further along.
Steal forAny repeating content format such as a weekly series or brand channel where visual consistency compounds over time
04:40list
Model Selection Heuristic
- GPT Image 2 for shots with on-screen text
- Nano Banana Pro for photorealistic shots without text
- Recraft for skin texture
- Video model selected per prompt automatically
Agent One picks the image and video model for each shot based on prompt content. The user never selects a model manually.
Steal forUnderstanding which model to reach for when not using an agent
12:36list
Cinematic Effects Stack
- Saturation boost via Color tab
- Edge Detect at low brightness via ResolveFX
- Horizontal and vertical scanlines with offset fade-out
- Dust overlay at 8 percent opacity Screen composite
- Camera zoom ease-in/out on adjustment clip
Five layered effects via adjustment clips that push DaVinci Resolve screen-content animations from flat to cinematic.
Steal forAny UI or screen-recording animation or b-roll sequence that needs a premium finish
16:10concept
Asynchronous Bounce Keyframe Pattern
Animate width and height bounce with the same decreasing-oscillation frame pattern (10, 9, 8, 7 frames between keyframes) but offset the height animation by two frames. The phase difference makes the widget feel organic rather than mechanical.
Steal forAny bouncy widget, pop-up card, or UI element animation in Fusion
How they asked for the click.
VERBAL ASK
22:20subscribe
“Make sure you subscribe so you don't miss that tutorial.”
Soft verbal outro pointing to a follow-up sound design video. Also an early mid-roll subscribe ask at 00:42.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
AFFILIATECommission earned if you click.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
