Wiring ChatGPT Into DaVinci Resolve to Auto-Edit an Entire Video
Danny Why runs GPT-6 Astra Ultra through DaVinci Resolve Studio's MCP server, escalating from a 15-second animated loop to a fully illustrated 12-minute talking-head edit.
Posted
1 weeks ago
Duration
Format
Demo
educational
Views
17.8K
236 likes
57 · 43
Big Idea
The argument in one line.
Connecting ChatGPT (running GPT-6 Astra Ultra) to DaVinci Resolve Studio through an MCP server lets it generate, animate, and place native motion graphics from a single prompt, and with the Higgsfield plugin it can auto-illustrate an entire unedited talking-head video in about 30 minutes.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
A YouTuber or editor who already pays for DaVinci Resolve Studio and wants to know if ChatGPT can realistically automate motion graphics work.
Someone curious what an MCP-connected AI agent can and can't do inside real editing software, not just standalone image generators.
A solo creator without a dedicated editor who wants a faster way to add illustrations to a talking-head video.
SKIP IF…
You're on DaVinci Resolve's free version. This entire workflow requires the paid Studio license plus an MCP server connection.
You need a finished, publish-ready edit. Every test in this video still needed a manual fix or review pass before it would be usable.
TL;DR
The full version, fast.
A creator connects ChatGPT (GPT-6 Astra Ultra) to DaVinci Resolve Studio through an MCP server and runs it through four escalating tests: a 15-second animated apartment scene, a purple gradient step diagram, a 3D warship battle with camera moves and sound, and finally a full 12-minute unedited talking-head video. Each motion graphic renders as an editable native Fusion composition rather than a flattened video file, and the Higgsfield MCP plugin supplies the actual illustration assets for the full-video test. Cost tracks against a weekly ChatGPT usage allowance that falls with every render, and the finished auto-edit lands close to, but not quite, publish-ready.
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.
Introduces the premise: ChatGPT connected to DaVinci Resolve can edit video and build motion graphics, provided you have the paid Studio version and an MCP connection.
00:31 – 01:59
02 · Setting up: GPT-6 Astra Ultra + DaVinci MCP
Shows the DaVinci MCP server registering with ChatGPT, checks the weekly usage allowance in billing settings, and reads out the first prompt before sending it.
01:59 – 03:35
03 · Test 1: an animated apartment illustration
A one-line prompt produces a 2D apartment scene with a subscribe end-card as an editable native Fusion composition; a follow-up prompt adds hand-typing and plant motion.
03:35 – 06:14
04 · Test 2: purple gradient step diagram
A curved-line 'Step One / Step Two / Step Three' motion graphic renders on a purple gradient background in about 10 minutes of AI working time.
06:14 – 08:15
05 · Test 3: two warships firing at night
A more complex 3D-camera battle animation with cannon flashes, smoke and sound renders from a single prompt in roughly 40 minutes, with one visual detail (bullet impact point) left for a follow-up fix.
08:15 – 09:59
06 · Installing the Higgsfield MCP plugin
Explains why this workflow is faster than other AI-editing complaints (DaVinci's dedicated MCP), then installs Higgsfield as a ChatGPT plugin to supply illustration assets.
09:59 – 15:14
07 · The real test: auto-illustrating a full 12-minute video
Drops in an unedited 12-minute talking-head video and asks ChatGPT plus Higgsfield to add matching illustrations throughout; reviews the finished split-screen edit, its cost, and its resemblance to established YouTube formats.
Atomic Insights
Lines worth screenshotting.
Connecting ChatGPT to DaVinci Resolve through an MCP server only works with the paid DaVinci Resolve Studio license, not the free version.
A single-sentence prompt can produce a fully editable native Fusion motion graphic in DaVinci Resolve, not just a flattened rendered clip.
The first draft of an AI-generated animation is often static; getting real motion usually takes one follow-up prompt.
AI-generated Fusion node graphs are dense enough that even the creator running the test calls them something no human editor would build by hand.
A 12-second, 3D camera-driven battle animation with sound and smoke rendered from one conversational prompt in about 40 minutes.
DaVinci Resolve has its own dedicated MCP integration, credited as the reason this workflow runs faster than other AI video-editing setups people complain about.
The Higgsfield MCP plugin is what actually generates the illustration images that ChatGPT then times to a video's audio.
Auto-illustrating an entire unedited 12-minute talking-head video, syncing new graphics to the narration throughout, took about 30 minutes.
A full-video auto-edit can burn an estimated 280 to 500 image-generation credits if it produces around 70 separate illustrations.
Without being asked to, the AI defaulted to a known-effective YouTube format: webcam on one side, a new graphic roughly every 15 seconds, each fading in and out on its own.
Weekly ChatGPT usage allowance dropped from 48% to 27% remaining across four escalating render tests in one session.
Even an AI-assisted edit with correct sync and appealing illustrations still reads as a first draft that needs a manual review pass.
Takeaway
What ChatGPT connected to DaVinci Resolve can (and can't) do yet
AI EDITING TEST
Wiring GPT-6 Astra Ultra into DaVinci Resolve Studio through an MCP server lets it generate, animate, and place native motion graphics untouched by hand, and even auto-illustrate a full 12-minute talking-head video, though every result still needed a manual note or two to land.
02Setting up: GPT-6 Astra Ultra + DaVinci MCP
The workflow only functions with DaVinci Resolve Studio, the paid tier, connected to ChatGPT through a dedicated MCP server; the free version has no equivalent hook.
Model choice affects output quality and speed: everything here ran on GPT-6 Astra set to 'Ultra', with a warning that a lower tier can produce slower or worse results.
Every generation draws down a weekly usage allowance visible in ChatGPT's billing settings, which fell from 48% to 45% remaining after just the first two-part render.
03Test 1: an animated apartment illustration
A single-sentence prompt (animated apartment scene, camera zoom-out, subscribe end card) produced a finished, editable native Fusion composition in about 20 minutes.
The first pass of a generated animation often comes back essentially static; a follow-up prompt asking for motion took a second render cycle to fix.
04Test 2: purple gradient step diagram
A more complex diagram request (purple gradient background, a curved line connecting three labeled steps) still rendered as a native, editable animation in roughly 10 minutes of AI working time.
AI-generated Fusion node graphs are dense enough that even the creator running the test repeatedly calls them something no human editor would build by hand.
05Test 3: two warships firing at night
A 3D camera-driven battle scene with cannon fire, smoke and sound rendered from one prompt in about 40 minutes, with a minor inaccuracy (bullets missing their target) left for a follow-up fix rather than blocking the result.
06Installing the Higgsfield MCP plugin
Higgsfield installs into ChatGPT as an MCP plugin in a few clicks and is what actually generates the illustration assets that DaVinci then times to a video's narration.
DaVinci Resolve's own dedicated MCP integration is credited as the reason this setup runs fast, compared to other people's reports that AI video editing is slow.
07The real test: auto-illustrating a full 12-minute video
Auto-illustrating an entire unedited 12-minute talking-head video, syncing new graphics to the narration throughout, took about 30 minutes and left roughly 27% of the weekly allowance remaining.
A rough cost estimate for the image side alone: around 70 separate illustrations at about 4 Higgsfield credits each works out to 280 to 500 credits spent purely on visuals for one video.
Without being asked to, the AI defaulted to a known-effective YouTube format: webcam on one side, a new graphic roughly every 15 seconds, each fading in and out on its own.
The finished auto-edit still needs a human review pass; it's presented as a fast first draft, not a publish-ready video.
Glossary
Terms worth knowing.
MCP (Model Context Protocol)
A connection standard that lets an AI model like ChatGPT directly control software such as DaVinci Resolve, rather than just generating text or files for a human to import.
DaVinci Resolve Studio
The paid tier of Blackmagic's DaVinci Resolve editing software, required here because only the Studio version exposes an MCP server the AI can connect to.
Fusion composition
DaVinci Resolve's node-based motion graphics and compositing system. AI-generated animations in this video render as native, editable Fusion node trees rather than flattened video clips.
Higgsfield
An AI image and video generation platform that plugs into ChatGPT via MCP, supplying the actual illustration assets used to auto-decorate a video.
GPT-6 Astra Ultra
The specific ChatGPT model and performance tier used throughout the video; the creator warns results may be slower or worse on a different model or tier.
“We're at the beginning of something that will completely change how video editors edit.”
bold standalone claim about AI's impact on editing work→ TikTok hook↗ Tweet quote
08:26
“What ChatGPT does, it's actually just control your computer.”
demystifies the whole mechanism in one line→ IG reel cold open↗ Tweet quote
13:57
“You're really just listing events. There's no tension, no conflict, and no reason for the viewer to care about what happens next.”
a sharp piece of writing-craft criticism embedded inside the demo itself→ newsletter pull-quote↗ Tweet quote
The Script
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.
17px
Okay, so if you didn't know, ChatGPT can now help you edit videos for you. In this video, I wanna show you how you can do the exact same thing, how you can connect ChatGPT to DaVinci and have ChatGPT edit videos for you.
Not only that, but I wanna test it out. I wanna see what's possible. What kind of motion graphics can we create?
So if you wanna learn how AI can help you edit videos, this video is for you. But first, you need to understand that at the moment, this only works with the paid version of DaVinci. And after you have the paid version of DaVinci, you have to connect to the MCP server.
Once you do this, you should have the same results I have. All right, before I start, I want to show you the prompt I'm going to be using. And I also want to show you how many credits I have.
So basically to check your credits, I think you just have to go over here to your account, go to settings. And over here where it says usage and billings, we can see that my plan is the pro plan, which is $100 a month. If we scroll down over here, we can see that my weekly usage limit is at 48 % left.
So I'm going to go back. I'm on Codex and I'm using GPT -6 Astra Ultra, right? So I have it set at Ultra.
If you're using something else and you don't have it set at Ultra, it might not work. It might be slower. You might get a different result.
Also, your computer matters over here because it's using your computer to create these graphics. Now, here's the prompt I'm going to be sending. and i'm basically asking for a 15 second animation that has like that specific 2d style for example iman gazi uses over here and i want the animation to be of a person inside an apartment the camera zooms out and then it says subscribe to the anyway because why not so i'm gonna send this prompt and that's it let's see what's going to happen I'm just going to leave my mouse over here and I'm going to speed up the video as much as possible.
But at the end, I'm going to show you how much time it took to create this video. So, for example, over here, you can see it's 5 p .m., five and two minutes. So at the end of this video, I'm going to show you the time so you know exactly how long it took to create this motion graphic.
And I'm also going to show you how many usage credits I have left. So let's speed up the video.
So I just came back to the computer and it seems that the video is ready. I have no idea when it's ready. I'm gonna have to check post -production.
Right now it's 5 and 22 p .m. And so let's go ahead and check out the video. I'm very curious.
This actually looks pretty good already. It's a little bit laggy because we're inside DaVinci. Now one thing I noticed, I don't like the subscribe to Danny Y over here font.
This is something I would go in and change myself. But the illustration itself, it's actually pretty amazing. I also don't like the fact that nothing over here is moving.
It looks more like a photo. But if I want, I can go back to ChatGPT and tell it to add some motion. And that's exactly what I'm going to do.
But before that, let's also check the usage. So how much did this cost to create? So if you go over here to settings and we check out billing, we can see...
that I have 45 % left. That's pretty okay in my opinion so far. Let's go back to the app and I'm gonna send a message to ChatGPT and ask it to add a little bit of motion.
Could you please add a little bit of motion to the 2D illustrations? For example, make him type on the keyboard or move the mouse. Just make it feel a little bit more dynamic.
So I'm gonna send this prompt. I'm gonna come back once it's ready.
All right. I'm back. ChatGPT seems to not be ready, but I do see that it rendered the second video, which I think means it's ready.
Right now it's 5 .31 PM. Let's check out the end result. Let me play it.
But let me actually play the rendered version because inside DaVinci, it's pretty laggy. So let's check out the rendered video.
So you can see over here, the hand it's typing on the keyboard. And honestly, this looks pretty good. The plant is moving a little bit.
In my opinion, it did a pretty good job. I think in total, it took like 20 minutes or something like that. Let's check the usage over here.
So for the weekly usage limit, we have 44 % left. Let's go ahead and create something else. But before we start a new project, I would like to check out the Fusion Nodes.
I'm curious how the Fusion Nodes look like. So wow, this looks crazy. This looks like made by a robot.
Like, come on. Tell me what human will create something like this. Just look at this.
Oh my God. oh my god but yeah let's go ahead and create something else now before i start again i want to show you how many credits i have left like how much percentage i have left so if you go over here the percentage we see we have 44 left now let's go back to chat gpt and here's what i want to create i want to create a motion graphic where the background is purple has beautiful gradients and there's a line that goes from the bottom left and in a beautiful curved way and it hits three dots that says step one step two step three Let's see how ChatGPT Astra Ultra can handle this.
You can see the time, it's 5 .50 p .m.
All right, so let's see. Wow, I just got back to my computer and it's literally rendering right now. It's at 46%.
And from what I'm seeing on the screen, I already like it. Now, as for the time, as you can see over here down below, it's 6 p .m., literally 6 p .m. Now, let's check out the animation.
I'm going to play it over here just for a second, just so you can see it's literally inside DaVinci. The line goes really nice. Oh my God, I actually like this animation.
Let's go to Fusion over here to check how the Fusion tab looks. Again, just a crazy Fusion tab, just like before. A human can't make this.
A human will not make Fusion nodes like this, okay? Now, let's check out the actual animation.
Yeah, I already like this. As I said, I don't like how small the dots are and how thin the line is, but overall, I really, really like this. I mean, it did a great job.
As I said, you can go back to ChatGPT and modify things. But let's go back to ChatGPT and see how much did it cost. So I'm going to go back here to my profile settings, and if I scroll down, 42 % left.
Now, let's go ahead and try to create something else. For this next animation, what I want to create is... two warships firing at each other.
I want the colors to be blue, as you can see this is brown over here, like gradient blue just like the first animation, and basically these two ships are firing at each other. I'm going to just leave the computer and leave my mouse over here and let ChatGPT do its thing. I will see you when this is ready.
All right, so I'm back. I have no idea how long this took. It's 7 .13 p .m.
as you can see over here. And from what I see on the screen, I'm actually impressed. This looks pretty good.
First, let's check out the final result over here in ChatGPT because this is the final render. Let me make full screen and let's play this. So we have two battleships.
I don't really like how... the freaking bullets like the freaking ammo hits the water but we can obviously fix this with chatgpt we can tell it that we want the bullets to hit the other ship and actually break it but i'm actually impressed with the quality like these ships have a lot of detail and they look pretty good like it's it's hard to tell this is ai slop right this is pretty interesting so all we have to do is just fix how the bullets land and obviously it's pretty easy to fix this I'm not going to go ahead and fix it because by now you should already understand that Chajupati is pretty much able to fix those things.
What I want to do instead is go into the fusion over here to see the nodes and how they look. So, oh my God, yeah. So look at the nodes.
This is not how a human is going to create the nodes. I've been saying this a lot, but come on, just look at this. Like you have to be crazy to think this is what a human will create.
Just look at these nodes. Let's go check. the chat gpt usage now because i want to see how many credits we have left how much percentage we have left so i'm going to go over here to settings usage and billing scroll down and we have 35 left that's actually pretty impressive as you can see the time is the same over here we have the time over here this is actually pretty impressive like i thought i'm gonna have like 30 left or 25 left i'm very impressed with this i mean we're at the beginning of something that will completely change how video editors edit now some people were saying that they tried it and it was slow but i don't think it's slow i think you're using the wrong model or you're using ChatGPT the wrong way.
Some people think that just by prompting ChatGPT to connect to some program, it's going to connect. But what ChatGPT does, it's actually just control your computer. The reason why it's so fast inside DaVinci is because DaVinci has a dedicated MCP, which makes things much faster.
That is why this only works with the DaVinci paid version. Now, here's what I want to do. I want to see if ChatGPT and Cloud can actually edit an entire video.
And what I mean by edit an entire video, I mean, I want to see if... I put in some footage. Let's say I put in 12 minute footage and all the graphics are going to be created by ChatGPT.
Let's see if that's possible. Now, here's what I want to do. I have this video of me that's 12 minutes long.
As you can see, all I do in this video is talk. It's just me talking and there's no footage. There's no visuals.
There are no graphics, no illustrations. I'm scrolling through the video as you can see, and there's no graphics, just me talking. And what I want to do is I'm going to drag and drop this video over here.
Basically, the video already has cuts. I already cut the video, but I don't have any graphics. And what I want is for ChatGPT to add all the graphics and illustrations in my video to also match the audio.
But for that, I'm going to be using Higgs field. And here's why. The reason why I want to use Hixfield is because with Hixfield, I can generate the images, the illustrations, and ChatGPT can just add them in DaVinci.
Now, if you want to try this yourself, there's going to be a link down in the description of this video to Hixfield. Once you click on that link, it should send you to the Hixfield dashboard over here. On the Hixfield dashboard, what I want you to do is click over here at the top where it says MCP.
Once you click on MCP, you can literally just click on this button over here which says add Higgs field plugin and it's going to automatically add the Higgs field plugin to your ChatGPT. Now for this to work, you need to be logged in on Higgs field and you also need to be logged in on ChatGPT.
That's it. Once you do this, you will have the Higgs field skill installed on your ChatGPT. Now once you have the skill installed, you can go back to DaVinci, you can open ChatGPT and now we can ask ChatGPT to edit this video.
And that's exactly what I'm going to be doing. So I'm going to send a vocal message and show you the exact prompt I'm sending because this is not even a prompt. I'm simply just asking.
Now, before I start, I do want you to see the time. It's 7 .20 p .m. And my usage credits right now, I'm going to check them once again.
They are at 35%, just so you know and you understand how much this costs and how much time this takes. I've created a new project inside DaVinci. It's called AI Test 5.
Now, I dropped in a video file of me that's 12 minutes long. Now, again, I'm using GPT -6 Astra Ultra. The time, it's 7 .23 p .m.
Over here, we have DaVinci and ChatGPT already started working. As I said, I'm gonna leave the computer and let ChatGPT edit my video.
All right, so... The render is almost ready and I'm so excited to see this. As you've seen, I haven't moved my mouse.
You can pause the video on the speed up and check ChatGPT working. I haven't done anything. Now, as you can see, ChatGPT is not ready, but I think the video is ready.
Now, the first thing I want to check out is the time. As you can see over here, it's 8 and 2 minutes. And if I remember correctly, I started this around 7 .30 p .m., which means adding footage to this entire video took 30 minutes.
Now, the second thing I want to check is how many credits we spent to add footage. So let's go over here, go to settings, usage and billing, and let's see how many credits we have left. We have 27 credits left, which honestly is pretty good because we just added footage.
to the entire video, but we did use Higgs field to create these images. Now I'm not sure how many images there are, but if there are 70 images, each image I think is like four credits. So it can be around 300 to 500 credits.
Now let's minimize the chat GPT window and let's check this out because I'm super curious. It does look promising. It cropped my video.
It added round corners over here and I'm on one side and on the other side I have all these graphics over here and you can see we have all the graphics over here created with Higgs field. I think if I go back to ChatGPT, we can see all the graphics here created with Higgs field. Let me zoom out on the timeline.
And as you can see, all the graphics are already added. And this literally took like 30 minutes, 40 minutes. That's just crazy.
Now, what I want to do is check out if the graphics fit. with my audio. So let me play the video for a little bit and see how it looks.
My last upload on YouTube is my best performing upload so far. As you can see, it's a 1 out of 10 upload. It even got those cute fireworks on it.
And the feeling you get when your video performs good on YouTube is amazing. So there's a graphic each 15 seconds, I think. I also like that they fade in and they fade out.
So I didn't ask ChatGPT to do that, but it automatically did it, which is pretty cool. Before I was able to achieve good results on YouTube, my videos were always getting stuck below 1 ,000 views. And if you're watching this video, I'm pretty sure you can relate to all of these stories.
This actually looks pretty good because I know a lot of YouTube channels that usually have this format where they have graphics or text on one side. and they have their webcam on the other side. So for example, Ed Lorenz over here, who has 103 ,000 subscribers, if we go to his channel and we check one video, we will see that he has this style of videos.
He basically has his webcam over here on one side, and over here he has text, or he has graphics, and a lot of big YouTubers actually use this formal style. which is why I think this is pretty cool because as you can see, the entire video has graphics over here on one side and it has my webcam over here. Now, let me just go over here at the middle, see how it looks over here in the middle.
If the graphics still sync with my audio, that's what I want to know. So I'm going to play it from here. But you're really just listing events.
There's no tension, no conflict, and no reason for the viewer to care about what happens next. Your video becomes very predictable. Your video becomes very predictable and the graphic says predictability.
but because this happened something went wrong therefore this person had to do something else but and this one also shows exactly what i'm talking about this happened but this happened therefore right so then that created another like for example even ali abdel over here with almost 7 million subscribers uses this style to make videos so you can see he has his webcam on one side and on the other side he shows graphics he shows videos he shows text this style seems to work And ChatGPT connected to DaVinci seems to be able to create this style pretty efficiently.
Now, if you want to learn how ChatGPT can help you edit videos inside CapCut, you might want to watch this video over here because inside this video, I'm going to show you how ChatGPT can edit videos for you inside CapCut. It can create motion graphics, it can add footage, and many more things. So if that sounds interesting and you're a CapCut user, you might want to click on this video over here.
The Hook
The bait, then the rug-pull.
ChatGPT can already reach into DaVinci Resolve Studio and build native motion graphics from a sentence, so this test just keeps escalating the ask, from a fifteen-second apartment loop to an entire illustrated 12-minute video that no human touched until it was almost done.
CTA Breakdown
How they asked for the click.
VERBAL ASK
14:52next-video
“if you want to learn how ChatGPT can help you edit videos inside CapCut, you might want to watch this video over here”
soft cross-promo to a companion video on the same AI-editing workflow using CapCut instead of DaVinci; the running CTA throughout is actually baked into the demo content itself via repeated 'Subscribe to Danny Why' text inside the AI-generated illustrations.
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
A screen-recorded walkthrough of turning a raw voiceover into a fully synced, AI-generated b-roll edit — using a timestamped transcript as the bridge between audio and images.
A screen-recorded demo of ChatGPT's Codex agent connected to DaVinci Resolve over MCP, cutting a raw talking-head recording and generating synced motion graphics with no manual editing.
A screen-recorded build of a full 3D motion graphic inside CapCut, using a Claude Code-generated asset and an image-to-video AI to carry it past the timeline's edge.
A 12-minute tutorial showing how Claude Code, Manim, and Higgsfield can replicate 3Blue1Brown-style math animations with the key prompt locked behind a $9/month community.
00:11
00:31
00:51
01:02
01:14
01:25
01:37
01:48
01:59
02:22
02:34
02:45
02:57
03:08
03:19
03:31
03:42
03:54
04:05
04:17
04:24
04:39
04:51
05:14
05:25
05:37
05:48
05:59
06:11
06:22
06:34
06:45
06:57
07:19
07:31
07:42
07:54
08:05
08:14
08:28
08:39
08:51
09:02
09:12
09:25
09:48
09:58
10:11
10:22
10:34
10:45
10:56
11:08
11:14
11:31
11:42
11:55
12:05
12:16
12:28
12:51
13:02
13:14
13:25
13:39
13:46
13:55
14:11
14:22
14:38
14:44
14:52
13:47
15:02
13:59
29:59
15:01
12:22