This DaVinci Resolve Plugin Automates Your Edits With Any AI Model
A VFX artist runs the AI Video Studio plugin through six escalating tests, from a simple AI transition to a full eight-minute automated edit, to see how much of a real edit Claude and three AI generation engines can actually carry inside a live DaVinci Resolve timeline.
Posted
2 days ago
Duration
Format
Review
educational
Views
4.3K
65 likes
57 · 43
Big Idea
The argument in one line.
A DaVinci Resolve plugin that wires Claude and three AI generation engines directly into the timeline can automate real editing tasks end to end, but every test still needed a human polish pass before it was publishable.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
A video editor or VFX artist curious whether an AI plugin inside DaVinci Resolve can carry real editing tasks, not just generate isolated clips.
Someone who already uses Claude or ChatGPT for text work and wants to see those same models writing prompts and making edit decisions inside a timeline.
A creator deciding between paying per generation on a cloud AI video API and running models for free locally through ComfyUI.
SKIP IF…
You're looking for a DaVinci Resolve editing tutorial for beginners. This assumes you already know the interface.
You don't have a GPU with at least 24GB of VRAM and have no interest in cloud generation costs. Most of the demo depends on one or the other.
TL;DR
The full version, fast.
AI Video Studio puts Claude, ChatGPT, and three AI generation engines, cloud Higgsfield and fal.ai or free local ComfyUI, directly inside a DaVinci Resolve timeline. An assistant panel writes prompts, picks models, and can act on the project itself: importing footage, cutting the best takes, color grading, building Fusion titles, and making word-synced captions. Across six escalating tests, from an AI start and end frame transition to a fully automated eight-minute edit from one prompt, the system handled the mechanical work but still needed manual timing fixes, cut corrections, and taste calls before anything was publishable.
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.
Kevin opens on camera with the pitch: build everything inside DaVinci Resolve using one tool, editing, captions, AI visuals and images, locally on your own GPU or on the cloud through the Higgsfield API, without leaving the timeline, replacing the need for ten separate apps and subscriptions. He notes the plugin already has over 1,000 editors using it.
01:33 – 02:20
02 · Install & interface tour
He shows where the plugin lives in DaVinci Resolve Studio, under Workspace, Workflow Integrations, AI Video Studio, and demonstrates pinning the panel to the top and docking it left, right, or centered depending on how much timeline space he needs.
In Settings, the engine dropdown switches between the Higgsfield API, fal.ai, and ComfyUI. ComfyUI is different from the other two: it renders locally and for free, as long as ComfyUI is installed and running in the background, with the plugin pointed at its server address.
03:37 – 05:02
04 · Free local AI images on your own GPU
He demonstrates image generation with a local Flux.2 Klein model: a text prompt plus full ComfyUI-style controls for seed, steps, and quality, along with camera and lens presets like a cinema camera, a specific lens, depth of field, and a 'golden hour' lighting look, all generating for free on his own GPU.
05:02 – 05:30
05 · Set up the AI Assistant
Back in Settings, he connects the Assistant to Claude under his own plan, no separate API billing, and shows ChatGPT and a free NVIDIA serverless API key as alternatives for anyone who doesn't want to pay for either.
05:30 – 07:03
06 · Let the Assistant set up your AI video
He opens the Assistant chat and asks for a simple image prompt, then asks it to build an actual AI Video Studio job directly: picking a model, filling in settings, and queuing the generation, demonstrated with a 'dog in a field' prompt and a local MiniMax run with a wide-angle lens and an FPV drone camera move.
07:03 – 07:27
07 · Cloud generation & real costs
Switching the engine to Higgsfield and then to Kling 3.0, he shows the actual dollar cost and credits a four-second cloud clip will use, about 50 cents in this example, as the tradeoff against free but slower local rendering.
07:27 – 08:16
08 · Claude Opus 5.5, the MCP & Assistant presets
For the more advanced tests, he switches the Assistant to Claude Opus 5.5 running through the MCP, and tours the preset library: saved prompts for repetitive jobs like organizing a timeline, building B-roll, color grading, audio design, and making animated captions.
08:16 – 10:12
09 · Test 1: AI transition with a start & end frame
First test: a smooth AI transition between a ground-level shot of a girl in a field and a top-down drone shot, using Seedance 2.5 with a start frame and an end frame. He frames it as how a VFX artist would approach the shot manually, now generated from one instruction, and it costs $3 to generate at 1080p.
10:12 – 11:44
10 · Polish it in Resolve: speed ramp & smooth cut
After dragging the AI transition onto the timeline, he manually speeds it up to 170%, adds a keyframe so the ramp eases in and out smoothly, then trims and adds a 'smooth cut' effect at both ends where the real footage meets the AI-generated footage.
11:44 – 13:16
11 · Test 2: Restyle a clip on your timeline
Second test: restyling an active green-screen clip of a guitarist using Seedance 2.5 and a saved 'volcano suit' character element, with a prompt that changes the performer's outfit and the environment to match a reference image while keeping the original movement. The 720p result is close enough to upscale rather than regenerate.
13:16 – 16:00
12 · Test 3: Import, organize & cut your best takes
Third test: a 'Setup From Folder' preset points the Assistant at a raw footage folder, and it imports the clips, identifies interviews, B-roll, and drone shots, keeps only the good takes by cutting silence and repeated lines, and converts the color so the footage is viewable. Kevin notes the cuts were mostly accurate but caught a few repeated lines it didn't recognize as mistakes, so it still needs human review.
16:00 – 17:15
13 · Test 4: AI color grade (LUT) in 1 minute
Fourth test: dropping in a reference image, or just asking for a 'teal and orange' look at medium strength, generates a usable LUT in about a minute, which he then copies across every clip in the timeline with Alt+V.
17:15 – 19:08
14 · Test 5: Motion graphics built in Fusion
Fifth test: motion graphics presets for lower-thirds, titles, and subscribe animations can render as a flattened clip from Claude, or get built natively inside Fusion as an editable composition, with individual layers, fonts, masks, borders, glows, and keyframe timing he can still hand-tune afterward.
19:08 – 21:00
15 · Test 6: Animated word-by-word captions
Sixth test: an animated caption preset highlights each spoken word, with controls for single-line layout, 9:16 aspect ratio, size, centering, and animation style. The generated captions pull from the project's own transcript and include at least one invented typo, 'venture resolves' instead of 'DaVinci Resolve', that needs a manual correction.
21:00 – 24:12
16 · Final test: AI visuals + a full edit in 8 minutes
Final test: from one written prompt naming a saved 'Camaro' element and a 'Swiss Alps' location, using Seedance 2.5 at 720p, the Assistant generates separate shots, builds a title animatic rendered outside Resolve, and assembles everything into a finished timeline with an end card, 'made by Kevin', in about eight minutes. Kevin then manually retimes several clips that ran too long.
24:12 – 24:52
17 · Verdict: what it can and can't do yet
Kevin calls the full-edit test a success given the time and prompt involved, but says he wouldn't hand it an entire project yet, only specific steps to speed up a workflow he still directs and checks himself.
24:52 – 25:22
18 · Where to get the plugin
He closes by pointing to the plugin on Gumroad as a one-time payment, like Resolve itself, and to his Skool community for people who want to stay current on AI editing workflows.
Atomic Insights
Lines worth screenshotting.
A DaVinci Resolve plugin now runs Claude, ChatGPT, or a free NVIDIA key as an in-timeline assistant that writes AI prompts and can act on the project directly.
The same plugin can generate AI video three ways: paid cloud APIs like Higgsfield and fal.ai, or 100% free on a local GPU through ComfyUI.
Rendering high-quality AI video locally needs at least 24 gigabytes of VRAM, and some local video models take up to 70 gigabytes of disk space.
A four-second AI video clip on the Higgsfield cloud API cost $3 in one test and 50 cents in another, depending on the model chosen.
The assistant can build a start-and-end-frame AI transition between two real shots from a one-sentence instruction, with no manual keyframing.
A 'Setup From Folder' preset can import raw footage, identify interviews versus B-roll versus drone shots, and cut only the good takes automatically.
The plugin can restyle an existing clip on the timeline; in one test it swapped a guitarist's wardrobe and background while keeping the original camera movement.
A full AI color grade, from a text prompt to an applied LUT, took about one minute in the demo.
Fusion-built motion graphics took about five minutes to generate but, unlike a flattened render, left every layer, font, mask, and keyframe editable afterward.
Word-by-word animated captions are generated from the project's own transcript, including at least one invented typo that needed a manual fix.
A single prompt naming two saved 'elements' and a camera style produced a full edited timeline, including a title card and end card, in about eight minutes.
Across six tests, nothing shipped without a manual pass: speed ramps, cut corrections, color notes, or re-timed clips were needed every time.
The creator's verdict: the plugin is a success for speeding up specific steps, but he would not trust it with a full project end to end yet.
Over 1,000 editors are already using the plugin, built and iterated on with direct community feedback.
Takeaway
AI can run the steps, not the project
WORKFLOW AUTOMATION
An AI assistant wired into a real editing timeline can reliably handle mechanical, well-defined steps, but every test still needed a human pass before the result was usable.
02Install & interface tour
The plugin panel lives under Workspace, Workflow Integrations in DaVinci Resolve Studio, not the free edition, so confirm you're on Studio before expecting to find it.
A floating AI panel that can pin, dock left or right, or center itself is worth building into any app you use constantly, so it never steals timeline space you need.
03Settings: Higgsfield, fal.ai & ComfyUI engines
Switching between a paid cloud API and a free local renderer should be a single dropdown, not a separate integration, if you want to actually use whichever one fits the job that day.
A tool that can render 100% locally and for free still has a real cost: it needs the host software running in the background and a GPU that can carry the model.
04Free local AI images on your own GPU
Free local generation is not a stripped-down mode: camera, lens, depth of field, and lighting presets can match a paid cloud tool's level of creative control.
A 'golden hour' lighting preset or a specific lens choice gets folded straight into the text prompt, a faster way to write detailed prompts than typing them from scratch every time.
05Set up the AI Assistant
Connecting an AI assistant through your existing Claude or ChatGPT plan, instead of a separate API key, removes a second subscription from the stack entirely.
A free API key option, even a basic one, is worth offering so a tool isn't gated behind a paid plan for anyone just trying it out.
06Let the Assistant set up your AI video
An assistant that can both write a prompt and fill in the generation settings itself removes a manual step: copying a suggested prompt somewhere else to actually use it.
Testing an assistant on something trivial, like 'a dog in a field,' is a fast way to confirm the automation sets up a real job before trusting it with the complicated version.
07Cloud generation & real costs
Showing the dollar cost per clip before generating it, 50 cents for a 4-second clip here, is the kind of transparency that keeps a pay-as-you-go AI tool from feeling like a trap.
Cloud generation isn't always faster than local: one cloud model took about seven minutes when local renders are normally quicker, so speed isn't a given just because it's paid.
08Claude Opus 5.5, the MCP & Assistant presets
Saved presets for repetitive jobs, organize a timeline, build B-roll, color grade, make captions, turn a one-off prompt into a reusable button.
The best model for a job and the cheapest model for a job are often different; picking a stronger model for the harder tests here was a deliberate tradeoff, not a default.
09Test 1: AI transition with a start & end frame
A smooth AI transition between two real shots can be generated from a start frame, an end frame, and one sentence, instead of hand-keyed motion.
Framing a test as 'how would I do this manually as a VFX artist' before asking the AI to do it is a good way to judge whether the output is actually comparable.
10Polish it in Resolve: speed ramp & smooth cut
An AI-generated clip rarely drops in at the right speed or cut point; a manual speed ramp, keyframed easing, and a smooth-cut effect at each seam is what actually sells the transition.
The join between real footage and AI-generated footage is the part worth the most manual attention, since that's exactly where a viewer's eye catches a mismatch.
11Test 2: Restyle a clip on your timeline
Restyling an existing clip while keeping its original camera movement needs a reference image of the new character alongside the prompt, not text alone.
A 720p AI result that matches the source footage's motion well is still worth upscaling rather than regenerating from scratch at a higher resolution.
12Test 3: Import, organize & cut your best takes
A single preset that imports footage, sorts it into interview, B-roll, and drone categories, and pre-cuts the best takes turns a tedious import pass into a one-line request.
AI-cut takes still repeated the same line twice in testing without catching it, so a first pass that claims to be done still needs a human watch-through before it's trusted.
Approving every automated change individually is available as an option, but skipping it and reviewing the batch result afterward is faster if you're willing to spot-check instead.
13Test 4: AI color grade (LUT) in 1 minute
Generating a usable color grade from a single word, like 'teal and orange,' or one reference image, compresses a grading pass into about a minute.
Copying one AI-generated LUT across an entire timeline with a single paste-attributes shortcut is what actually makes a one-minute grade useful at scale, not just on one clip.
14Test 5: Motion graphics built in Fusion
A flattened AI-rendered title looks identical to one built natively in the host app's compositor, but only the native version keeps every layer, font, and mask editable afterward.
Building a motion graphic in your own compositor instead of importing a rendered video is the difference between a title you can tweak and one you have to regenerate from scratch.
15Test 6: Animated word-by-word captions
Word-by-word animated captions generated from a project's own transcript still need a typo check; the AI will occasionally invent a word, like a misspelled product name.
Saving a caption style as a reusable preset is what turns a one-time setup into a standing tool you reach for on every future edit.
16Final test: AI visuals + a full edit in 8 minutes
A full edit assembled from one prompt naming two saved reference elements and a style, with no further input, took about eight minutes end to end.
Even a 'finished' AI-assembled timeline needs a manual pass on pacing, since individual clips can run longer than the cut actually needs.
17Verdict: what it can and can't do yet
Calling a result a 'success' and trusting it with a full, unsupervised project are two different judgments; draw that line explicitly rather than letting one imply the other.
The honest use case for a tool like this is speeding up specific steps inside a workflow you still direct, not replacing the decisions about what the final edit should be.
Glossary
Terms worth knowing.
MCP
Model Context Protocol, the connection that lets an AI assistant like Claude or ChatGPT act inside a third-party app, here running tools on a DaVinci Resolve timeline through your existing AI subscription instead of a separate API key.
ComfyUI
A free, open-source visual tool for running AI image and video models locally on your own GPU, with no per-generation cost once it's installed and running.
VRAM
Video memory on a graphics card. High-quality local AI video generation in this plugin needs at least 24 gigabytes of it.
LUT
Look-Up Table, a preset of color values applied to footage to create a consistent grade or style, like a teal-and-orange look.
Fusion
DaVinci Resolve's built-in node-based compositing and motion graphics page, used here to build editable animated titles instead of importing a flattened video render.
Start/end frame transition
An AI video generation technique where you supply the first and last frame of a shot plus a prompt, and the model generates the motion that connects them.
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
metaphor
What if you could build everything inside DaVinci Resolve using just one tool? Edit your long -form clips, make customized captions, or even make visual effects, AI videos, and images locally on your own GPU. Or also on the cloud using the latest Higgs field API for cheap generations, pay -as -you -go, but also just fall AI.
and all of this without ever needing to leave your timeline it's all in one place your edit the visuals everything we can drag and drop instantly and it removes the need for 10 different apps or subscriptions for your creations with the latest davinci resolve update we can now use cloth and codecs with your own plan using the mcp you can let the ai assist and work inside your davinci resolve project as an editor and creative this can actually help you problem solve or even fully automate tasks for you so you have more time to spend on other parts of the workflow we have over 1000 editors now using the plugin which is absolutely amazing to see and i hear your feedback directly it is constantly improving and new ai models land very fast if you're interested link is in the description now to keep it real i will test this from very easy to very hard tasks in different stages and of course it's still up to the artist to fine -tune the edits and make it a polished visual so let's set it up and start with level one here we are in davinci resolve and if you installed the studio version you can go to workspace workflow integrations and ai video studio
And here are all the apps we can choose from. There are many different types of tools, but I will go through it and showcase how the plugin works. For the window itself, we can pin it on the top.
So that way, when we click and resolve, it will always stay visible. then we also have a window layout so we can dock it to the right we can dock it to the left or have it centered like this so when we open up the node studio for example we have a lot more space when we are working in this interface but for now we can dock it to the left and go to our settings On the top right here, we have our settings and this is the brain of our plugin.
Here we can choose which mode it is in, your API keys and all sorts of things that you would like to use. Here is our engine where the AI is running. Right now it's using the Higgs field API.
So if you paste our API key from Higgs field, then we have options to use. all the video and image models that Higgs Field is using. But if you're using Fall AI, for example, you can switch it to Fall AI and then paste your API key in here.
You also have the option for Comfy UI, and this works a little bit different because with Comfy UI, we can also render locally so we can render AI 100 % for free on our own system. And for that, the only thing you need is a working version of ConfUI installed. Have it running in the background.
And right now it's using our server address. So if we scroll down, we can see all the different AI models that we can download and use on our own system. Keep in mind, some of these models take some space on your computer.
So especially the video models can be up to 70 gigabytes. And just to demonstrate, if you now go to image generation and then use an image model like Flux2Cline, for example, I can paste in a random prompt and then set up our settings. We can change the seat, the steps, all the control you will find in CompVUI, set the quality to max.
And here we have also camera and lens control. So we have all sorts of presets for the look of our images and our videos. For example, I can select a cinema camera and choose a master prime with 35 millimeter lens.
You can also change the shadow depth of field and our lighting. For example, I want a golden hour look, so I can select that and it will include it in the prompts. So right now when we generate, we can generate images for free on our own system.
Keep in mind, you do need a good graphics card for this. Otherwise, it will not work. Especially if you want to render high quality AI videos, then you need at least 24 gigabytes of VRAM.
And there we go. Now it's created an image like this. And we can preview this.
We can also drag onto our timeline and start using it. And of course, there's much more control we have. We can also add reference images from our timeline or even add your own characters and elements, just like we have in Higgs field.
Back in our settings, we can also set up our assistant. So right now we are using Claude, but you can also use ChatGPT and GPT -6. And all the models are available.
This is directly connected to your account. So you don't need any other API costs or anything like that. It's just your own plan.
If you don't want to use any of that, we can also go to NVIDIA and use serverless API key, which is a free API key with zero costs. And we can also use that. You can go to our assistant, which is this small button right here.
And this is where the power comes in. Here we can ask for prompts so you don't have to do it yourself. And this is the automation part of our plugin.
For example, I can write something very simple, like write me a prompt for a dog in a field. And here it comes up with a prompt that we can copy and paste. But we can also ask something like make it in the video generation in AI Video Studio.
And then it will set it up directly in your app. And here it will set up all the settings for your AI video that you want to create along with the prompt. And as you can see, it's already set up right here.
and if you want we can change the model we are still running the local models for example we can choose something like minimax h3 we can still change our lens if we want to for example i will choose a 14 millimeter wide angle lens and in the movement we can say something like an fpv drone shot we only need four seconds and then hit generate Then we have a 4 -second shot of an FPV drone with a dock in a field, all local and free.
But keep in mind, this will take a long time to render, especially this model. LTX 2 .5 or 1 will render much faster than Minimax. But we can also change it back to Higgs field, for example.
And here we can render on the cloud using a Higgs field API. And now if you go back to video generation and change it to something like cling 3 .0, then we can see the costs and the credits that it will take for a clip of four seconds. In this example, it takes 50 cents to create this clip.
But we can do much more advanced stuff like this. For example, I can go back to our assistant. And for this, I will use Cloud and Opus 5 .5 to get the best prompts and also using the MCP server.
And in the assistant, there's all sorts of things we can do. We can ask the chat directly what it needs to do in the resolve or generate. We also have presets.
So for repetitive tasks, we can create our own presets or use the ones we have right here. For projects, we can set up or organize our timeline. We can make B -roll.
We can cut or clip. We can even make animated captions in Fusion or color grade and make audio design. There's a lot of things we can do with the Assistant.
For the first test, here is some footage of a girl in a field. And I want to make a transition between this shot on the ground and then our drone shot. The way I would do this as a VFX artist using AI, I will do this with a start frame and an end frame.
The start frame will be from our first shot and then the end, our drone shot. So in the assistant, I will go to chat mode and create a new chat. And here I will set my first frame.
which is gonna be this one right here. And then our end frame will be this one. And with this icon here, I will just grab a still.
So we have this still in our assistant and we can use this for the start and end frame. Here's what I said to the assistant, make a smooth and fast flying transition between the first shot of the girl in a field and the second top down drone shot. Start an end frame transition using C dance 2 .5.
I will hit send. Let's see what the assistant will come up with. Here it will give us a card using Higgs field.
This is going to be $3 to generate with 1080p. It's using the image to video model, start and end frame with a duration of 4 seconds. If you want we can also select the audio on or use our presets but in this case we don't really need that.
So I will hit generate and let's see what type of transition it can make. For some reason this generation took about seven minutes. Normally using fall or comfy it's a lot faster but I don't know why but Hicksfield takes a bit longer to do that.
Now if we scroll up here is the results. It created a smooth transition from our first frame and our end frame. just like that and now we can drag this into our timeline just like so and here we have the final transition and since we are in this edit page right now i will do some minor tweaks first thing i will do is adjusting the speed of the transition itself so it's a bit faster so i will hit ctrl r on the keyboard for our speed handle and change it to 170 percent something like this i will also go to the keyframes and just add a keyframe for the speed so it goes from 130 20 to 170 and with this icon right here we see our curve we can smooth out that curve by going to easy in and out adjusting the handle so it's a smooth time remapping so now we have something like this But the transition between our real footage and the AI footage needs to be tweaked a bit.
So I will trim the clip and delete it just like so. Then going to the effects panel and adding a smooth cut effect. And this will just smooth that out and make sure it's a smooth transition just like that.
And same thing for the ending. And here is the final transition. Pretty cool, right?
It's pretty crazy how fast we made this work just by asking the assistant and Claude will make the prompts and make the generations using our plugin.
And the power of Resolve is that we can also restyle an active clip that is on our timeline. Here we have a green screenshot and I want to restyle this video so we can go to edit video. And here it will play the exact same clip we have in our active Resolve timeline.
And we have all sorts of presets here. We can draw the video, draw something and change it. Then we also have reframing, relighting and all sorts of things that are handy when editing your videos.
Now to demonstrate, we can use Seedance 2 .5, for example, which is the best capable model. We can type in a simple prompt like change the guitar player. Then we can use an element that we created, for example, this volcano suit.
And this is based off our elements. So we can choose this volcano suit guy. So here's the final prompt, change the guitar player to volcanic suit and change the environment based on reference image too.
Make sure the bitrate is set at high. And then I will generate our video. And as you can see, after doing that, it generated a video like this.
We can also add it onto our timeline. And if you place it side by side, you can see it matches the movement very well. And it's looking pretty good.
But this is a 720p clip, so we can still upscale this or use the AI superscale from DaVinci Resolve.
We can also do something way more simple. For example, here I made an intro for a YouTube video I was going to make. And I just want DaVinci Resolve to upload the footage, convert it, and pre -cut what is being said in the video.
So go to the Assistant. Here we have a preset called Setup from Folder. So here we can simply go to the folder of my footage.
And here it will identify interviews, B -roll shots, drone, or anything else. and we can select silence and takes which will keep only the good takes without any pauses and we can also choose to clean up the audio or anything else that we want for the assistant then it will start importing our footage making the folders and start cutting our clip With every step it's building, it will say it in the assistant as well.
So you can see exactly what is being done. If you want to have more control of the changes it will make, we also have a button for it. So every change it will make, you will need to approve it in the assistant itself.
But I will just keep it off. So it will just generate and create what it needs to do. Now it's created a folder for us.
the footage and already made a pre -selection of what is being said in the intro. But I noticed that the colors are not converted yet. So I can simply go to our assistant and say, vert this tree footage.
And now it's converted the footage in a viewable format. And let's see how the cut was made and if it's actually usable. I will give you my workflow inside of DaVinci Resolve that turns a basic 3D render of your timeline.
So here's how it works. So here's one of the mistakes it can make. I repeat myself here.
So that's something the AI doesn't pick up. So I'll just delete that. so here's how it works first we use real in most air and this looks all good from that animation and make it look good i'm doing this i'm i'm doing this here same thing i repeat myself so i'll cut this out and i corrected itself a few times but other than that all the cuts were actually pretty accurate and straight to the point and this took about one minute each to create this so I mean, this really speeds up my workflow and it's just a faster way of going through lots of footage.
But I will not rely on it fully yet because as you can see, it can still mess things up or not exactly take the best takes. So that is still something you need to supervise and check yourself. Now for color grading, you can also have a preset, for example, drop in a reference picture, and then it will create a LUT for you.
But we can also have a quick LUT, for example, a teal and orange look. I will leave the strength medium and then hit run. And now it will create a quick LUT for us that we can use for our video.
And again, after about one minute, it's created a look for us. This is the before, this is the after. I just made a quick orange and teal LUT, and we can still add another note, add some contrast, or even add a film look.
And there we go, in about one minute, I just created a cool looking look. And this took no effort at all. You can also add this LUT to all the clips so I can copy and then Alt V to paste all the attributes onto this clip as well.
So now everything is with our new LUT.
Then we also have presets for motion design. So if I scale this up, we have motion design presets right here. for lower thirds for titles for numbers for your subscribe animation like this and we can design it right here using cloth or describe it based off references from links or images and then this will go to cloud and generate an animation for you And there's two ways we can do this.
We can make the title animation, for example, using Claude. So we can render it right here. Then we can add this to our timeline.
And as you can see, this is a fixed motion graphic. But if we change it to build it inside of Resolve, then it will send a preset prompt in our chat and then build it using Fusion. and after waiting about five minutes it analyzed all our requests and it made a fusion composition for us and if you look at it it looks almost exactly the same as the one cloud made for us but the beauty about this is that we can go into our fusion page now and we can edit our individual layers for example we can select this work right here we can change the font we can also go to this mask right here You can change the border width and even the color if you wanted to.
So we can customize this. We can even add another note to this, for example, a glow. So we can add a glow to this line right here.
And if you know Fusion well, we can even go to our keyframes, this text right here. You can change the timing of this so it comes in later. And obviously, this is just a simple example.
In the timeline, I made this real. And let's say I want to have some animated nice captions that highlights word per word what I'm saying in the video. Easiest way to do that now is going to the animated caption preset.
And here we can design the preset that we want. There's all sorts of styles we can choose from. But the one I will use for this is the podcast preset.
This highlights every word that has been said. I can choose to have it on one line. I can also select aspect ratio 9 by 16.
And we can also change the size and all of the controls you will need. And I will also center it in the middle like so. You can also choose the animation, how fast it goes in and which type.
in this case i will do a simple fade animation and when everything is good we can select make captions and that will create all the captions on our active clip and as you can see it creates another track for us and if we preview this i will give you my secret weapon inside of DaVinci Resolve that lets you generate AI videos and visual effects.
And it created our captions. And this is based on our transcripts. But there's a few things we can change.
For example, DaVinci Resolve, it created venture resolves. That's not right. We can go into our captions and then change that typo.
So that corrects the word. And now it's DaVinci Resolve instead of that typo. and right here we don't need captions so i will just remove them and that's how i create animated captions for my reels and again you can also create your preset if you like this style we can save as preset and then save it right here then if you go to yours you will see that caption preset that you can always use whenever you want Now let's do something more challenging where we ask the assistant to make AI visuals, then place them in the timeline and create a cool edit for us.
Just a polished edit all from the assistant itself. Now for the final test, I will let our plugin generate AI visuals and then create an edit within our timeline. For this, we can either use our presets or just explain exactly using your references.
So I can say something, make a 10 second cinematic sequence of, and then hit an add. Then we can add a element that we already created. So here we have a car that we can use as an element.
So I will select the car as the object and that will add all the references in our chat and then say inside our world and then add Swiss Alps as location. just for this example here is the final prompt make a 10 second cinematic sequence of camaro which is the element using swiss alps which is another element as the location reference make the shots separately with all the prompts and elements input correctly using cdance 2 .5 720p after that create a title animatic as a clip rendered outside resolve with the title camaro Fast cuts and fast -paced shots.
At the end, I want an end card saying, made by Kevin. I literally have no idea what it's gonna create. You can go much more in detail and Claude will do the rest.
But just from this simple prompt, let's see what it's gonna create. Now, after a few minutes, it generated this card and it's now making the videos. Simultaneously, it also created this animation from the text.
I'll also correct it and say, when you are done, place all of the assets in a new timeline and create the final edit. now all the results are now in the media pool and it gave us a brief of what it did it created the edit and the end card and all in all this took about eight minutes to generate let's see the results in our timeline i haven't watched it yet it's just straight from the prompt and the instructions we gave it and let's see what it came up with so it has smooth title okay that's a bit weird timing is weird but overall that last shot is quite impressive all right a couple of things I would tweak the timing of these clips can be a bit better so as you can see these are very long clips so I will change it slightly now here's the same sequence i changed the timing a bit and also our animation and this is what we have now and all in all it's not that bad considering the time this took and also the prompt i gave it so i would say this is a success and you can definitely speed up
certain aspects of your workflow but i wouldn't do it for the full project i would use it for some aspects of it to speed it up but again these are just tools and we can use it to speed up our workflows or just make things faster all of these automations we can really fine tune it by making your own presets and own prompts and see what works what doesn't and there's a lot of customization you can add to your workflow plugin itself is now available on gumroad as a one -time payment, just like Resolve.
Now, if you want to learn more about DaVinci Resolve and Fusion and visual effects as a whole, we have a community on school if you want to stay up to date of the latest AI workflows. If you have any questions, make sure to comment down below and start a conversation. Thank you guys so much for watching and see you on the next video.
Peace.
The Hook
The bait, then the rug-pull.
Kevin opens with the pitch: edit, caption, grade, and generate AI video and images without ever leaving the DaVinci Resolve timeline, free on your own GPU or paid on the cloud. He spends the rest of the video testing whether that claim actually holds up across six escalating tasks.
Frameworks
Named ideas worth stealing.
08:16concept
Start/end-frame AI transition
Supply the first frame of one shot and the last frame of the next, add a one-line prompt, and an image-to-video model generates the connecting motion instead of hand-animating a transition.
Steal forany edit joining two unrelated shots without a traditional cut or wipe
13:16concept
Setup From Folder footage triage
point the assistant at a raw footage folder
it identifies interview, B-roll, and drone shots
it keeps only takes without pauses or repeated lines
it converts color to a viewable format
A single preset that imports, sorts, and pre-cuts an entire footage drop in about a minute per clip, before any human touches the timeline.
Steal forany editor with a large unsorted footage dump to triage before the real edit starts
CTA Breakdown
How they asked for the click.
VERBAL ASK
25:00product
“the plugin itself is now available on gumroad as a one-time payment, just like Resolve... we have a community on school if you want to stay up to date”
soft close after the verdict, one-time-purchase framing pegged against Resolve's own pricing model, plus a free community link for ongoing updates
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 one-time-purchase Resolve plugin runs Minimax, Seedance, and GPT image models locally or in the cloud, straight from the timeline, no subscription required.
A native DaVinci Resolve plugin puts fal.ai, Comfy Cloud, and free local rendering on the timeline, and gets weighed line-by-line against a $130-a-month platform.
A walkthrough of AI Video Studio, a DaVinci Resolve plugin that lets editors generate video, images, backdrops and 3D models without ever leaving the edit.
A 40-minute raw screen session where the only human work is a few typed prompts, and Claude Code does the cut, grade, graphics and sound inside DaVinci Resolve, After Effects and Blender.
A DaVinci Resolve editor connects an AI agent straight into the open project, then spends the video testing whether it survives real revisions, not just a first draft.
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.
00:23
00:47
01:06
01:44
02:03
02:41
03:00
03:19
03:57
04:16
04:35
04:45
05:51
06:10
06:30
06:49
07:46
08:24
08:43
09:02
09:21
09:40
09:59
10:37
10:56
11:15
11:40
12:12
12:31
12:50
13:09
13:28
14:06
14:25
14:44
15:03
15:22
15:41
16:19
16:38
16:57
17:35
17:54
18:13
18:32
18:51
19:30
19:49
20:08
20:27
20:46
21:24
21:43
22:02
22:21
22:40
22:59
23:18
23:37
23:56
24:34
25:13
16:54
17:21
05:42
40:45
19:10
15:14