This DaVinci Resolve Plugin Just Replaced Every AI Video Platform
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.
Posted
4 weeks ago
Duration
Format
Demo
educational
Views
90.4K
2.8K likes
57 · 43
Big Idea
The argument in one line.
A one-time-purchase DaVinci Resolve plugin that renders AI video locally or pay-per-generation in the cloud beats an all-in-one subscription platform, because subscription credits expire monthly while owned compute and pay-as-you-go generation never do.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You edit video in DaVinci Resolve and want AI generation, upscaling, or compositing without leaving the timeline.
You already pay for a subscription AI video tool and want to see what a one-time or pay-per-generation alternative actually costs.
You have a capable GPU (24GB+ VRAM) and want to generate AI video for $0 beyond electricity.
You do VFX or compositing work and want AI-assisted cleanplates, object swaps, or transitions instead of manual frame-by-frame work.
SKIP IF…
You edit in Premiere, Final Cut, or another NLE — this plugin is Resolve-only.
You don't have a GPU with at least 24GB VRAM and aren't willing to pay per cloud generation, since the free-local path won't be usable.
TL;DR
The full version, fast.
AI Video Studio is a paid, one-time DaVinci Resolve plugin that puts video generation, editing, and compositing models directly into the timeline, rendered through fal.ai (pay per clip), Comfy Cloud (credits), or a local GPU (free). Compared to Higgsfield's $99-130/month subscription with 90-day expiring credits, the plugin's pay-as-you-go and $0 local options only charge for what gets rendered. The video walks through real per-clip costs ($0.30-$3 in the cloud, $0 locally), then demos AI transitions, object composites, cleanplates, and restyling, all generated inside Resolve and dragged straight into the edit.
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 previews using every leading AI video, image, and 3D model directly inside DaVinci Resolve's timeline via a community-built plugin.
01:30 – 02:35
02 · Interface tour
Walkthrough of the AI Video Studio panel: video generation, edit video, image generation, backdrop generator, 3D model tools, and the node-based Node Studio for wiring reference images and prompts.
02:35 – 05:27
03 · Pricing vs Higgsfield
Breaks down the plugin's three render modes (fal.ai pay-per-generation, Comfy Cloud, local GPU) against Higgsfield's flat $99-130/month subscription with expiring credits.
05:27 – 07:03
04 · Real costs
Concrete per-clip pricing across fal.ai, Comfy Cloud, and local rendering, from $0.30 for a short 720p clip up to several dollars for max-quality 4K output, versus $0 on local hardware.
07:03 – 11:04
05 · AI transitions (free & local)
Live demo generating a locally-rendered AI transition between two real car shots using Minimax H3 in Node Studio, then blending it into the timeline with an offset cut and smooth-cut effect.
11:04 – 13:49
06 · Edit videos
Demonstrates compositing a new car into existing footage using masked reference images, a two-step image-then-video generation workflow, and multiple video-to-video models (Seedance 2.0/2.5).
13:49 – 15:42
07 · Cleanplates with AI
Uses AI video-to-video generation to remove cars from a road and add distant mountains to the background, then composites the result back over the original footage using Fusion masks.
15:42 – 16:30
08 · Restyle video
Restyles a dance performance clip locally with Minimax H3 by feeding it reference images of a new look alongside the original video and prompt.
16:30 – 16:54
09 · Where to get it
Points viewers to the Gumroad one-time purchase and the Skool community/course that teaches the underlying workflows.
Atomic Insights
Lines worth screenshotting.
A flat AI subscription charges the same in a slow month as a heavy one, so intermittent use quietly wastes money every billing cycle.
Higgsfield credits expire and never accumulate, so even a paid top-up can go to waste if it isn't spent before the cycle resets.
Pay-per-generation cloud pricing (roughly $0.30 to $3 a clip) only bills for output actually generated, which favors bursty or seasonal workloads over constant use.
Running open AI video models on your own GPU costs $0 per render, with electricity as the only marginal cost, but needs at least 24GB of VRAM to be practical.
A 5-second 720p AI clip can cost around $0.30-$1, but pushing to 4K at 30 seconds multiplies that cost, since resolution and duration are the two biggest price levers.
Comfy Cloud's credit system holds value for a full year on top-ups, versus 90 days for Higgsfield, which matters more the less frequently you generate.
Negative prompting (explicitly stating 'no cuts, no dissolve, no cutaways') controls unwanted artifacts in AI-generated video transitions better than only describing what you want.
Blending an AI-generated clip into real footage looks more seamless when you offset the cut point by a few frames and add a smooth-cut effect, rather than hard-cutting between the two.
Generating a reference image before generating a full video is significantly cheaper, so front-loading image generation catches composition problems before paying for a video render.
Feeding an AI edit model a masked reference frame plus an explicit instruction to preserve one detail (like a license plate) keeps it from inventing unwanted changes elsewhere in the shot.
Setting a composite blend mode to 'difference' is a fast way to visually verify how closely an AI-generated replacement clip tracks the original camera movement.
AI-based cleanplate work (removing objects, adding new background elements) can replace hours of manual VFX painting, but still benefits from manual masking afterward to preserve as much real footage as possible.
Restyling a clip works better by feeding an AI model reference images of the new look plus the original video and prompt together, not by prompting from text alone.
Takeaway
Owning your GPU beats renting AI video
COST VS OWNERSHIP
Subscription AI video pricing charges a flat rate whether you use it or not, while pay-per-generation and local-GPU rendering only cost something when you actually generate a shot.
02Interface tour
A model comparison panel that shows multiple video-generation models side by side removes the guesswork of picking one blind.
Node-based interfaces let you wire reference images, camera angle, and lighting into a single generation instead of describing everything in one text prompt.
03Pricing vs Higgsfield
A flat AI subscription charges the same in a slow month as a heavy one, so intermittent use quietly wastes money every billing cycle.
Subscription credits that expire and never accumulate mean even a paid top-up can go to waste if it isn't spent before the cycle resets.
Pay-per-generation pricing, roughly $0.30 to $3 a clip, only bills for output actually generated, which rewards bursty or seasonal use over constant use.
04Real costs
A 5-second 720p AI clip commonly costs $0.30 to $1, but pushing to 4K and 30 seconds multiplies that cost fast, since resolution and duration are the two biggest price levers.
Credit-based cloud rendering can run $20 to $100 a month depending on model choice, but unused credits there can hold value for a year instead of expiring in 90 days.
Running the same open models on your own GPU costs nothing per render beyond electricity, but needs real VRAM (24GB or more) to be practical.
05AI transitions (free & local)
AI-generated transitions between two real shots work best when you explicitly state what should NOT happen (no cuts, no dissolve, no cutaways), since negative prompting controls unwanted artifacts better than positive description alone.
Blending an AI-generated clip into real footage looks more seamless when you offset the cut point by a few frames and add a smooth-cut effect, rather than hard-cutting between the two.
06Edit videos
Compositing a new object into existing footage works better as two steps: generate a reference image first, then animate it, since image generation is far cheaper than video generation.
Providing a masked reference frame plus an explicit instruction to preserve one specific detail (like a license plate) keeps an AI edit from inventing unwanted changes elsewhere in the shot.
07Cleanplates with AI
AI-based object removal and background replacement can replace hours of manual VFX painting, but still benefits from manual masking afterward to preserve as much of the real footage as possible.
Setting a composite blend mode to 'difference' is a fast way to visually check how closely an AI-generated replacement clip tracks the original camera movement.
08Restyle video
Restyling a clip works better by feeding an AI model reference images of the new look plus the original video and prompt together, rather than prompting from text alone.
Glossary
Terms worth knowing.
fal.ai
A cloud API service that hosts many different AI generation models and bills per generation rather than a flat subscription.
Comfy Cloud
A hosted version of ComfyUI that runs AI generation workflows on remote GPUs using a purchased credit balance.
ComfyUI
A free, node-based interface for running AI image and video generation models, commonly self-hosted on a local GPU.
Node Studio
The plugin's node-based workspace for wiring together reference images, prompts, camera settings, and a chosen AI model into one generation.
Cleanplate
A version of a shot with unwanted objects removed, traditionally painted out by hand in VFX and here generated by an AI video-to-video model.
VRAM
The dedicated memory on a graphics card; local AI video generation needs a large amount of it (24GB or more) to run usable models.
Video-to-video model
An AI model that takes an existing video clip as input and outputs a modified version of it, used here for object removal, restyling, and edits.
“With AI Video Studio, it's a one-time payment, no subscriptions ever, and it lives right inside of DaVinci Resolve and connects really well to your timeline.”
core pitch line, states the whole value prop in one sentence→ TikTok hook↗ Tweet quote
06:42
“For free, zero dollars. The only thing you need to pay is of course your GPU and the electricity bill.”
“Even if you have a quiet month that you don't use it, you will still need to pay that monthly subscription.”
names the exact subscription pain point→ newsletter pull-quote↗ Tweet quote
13:13
“It's crazy what AI can do nowadays.”
short reaction line that works as an overlay caption on the composite demo→ TikTok hook↗ 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
metaphor
I will give you all my secret weapon inside of DaVinci Resolve that lets you generate AI videos and visual effects without ever needing to leave your timeline. We can use the best AI models both on the cloud and locally 100 % for free on your own GPU. Every leading model for videos, images, 3D all lives inside of our DaVinci Resolve software.
Together with the community we built a fully working plugin and it's not another browser or expensive subscription where you have to leave, resolve, download, re -import, none of that. We can edit our video, relight it, reframe it, upscale it, all in one panel. You can drag and drop it straight into your edit, no exporting, no re -importing.
There's also an advanced node -based UI, we can feed it with reference images, wire up a model and get a cinematic shot back without figuring out the right prompt for an hour. It's built by actual editors and VFX artists who are already using it.
With the community behind it, it keeps getting better and your support has been massive in improving this tool. When new AI models come out, it will land very fast. We even built an AI assistant that runs for free with a serverless API from Nvidia.
Honestly, this is the workflow that supercharges DaVinci Resolve and you can see it as your personal VFX or CGI assistant. Let me show you how it works and I will also give you some tips and tricks how to do AI visual effects. Here we are in DaVinci Resolve with our plugin opened up.
And as you can see, we have different tools. And the most common tool is video generation with AI. So here we have all the latest models we can choose from.
We can also change the aspect ratio, the duration. your resolution and we also have a camera and lighting option so you can choose which camera look you want and the lighting of our environment then we also have a preset for the animation and that way when you want to generate a video you don't have to look for the right prompts you also have an advanced workflow called node studio here we can generate with a node based setup so we can add prompts we can add reference images videos all in a node based system and here we can also change the options for the models in the settings we have different modes for example we have fall ai then also comfy ui you can also select our output folder so every video and image will be stored within that folder we can change that per project Now many of you have asked how the payment system works within the plugin itself.
I'm going to clarify this and show you how it compares to something like Higgsfield for example. This works for both the free version and the studio version of Davinci. Then we have three different modes we can choose from.
We have fall ai with every leading model you can think of within the plugin we can set up those api keys then we also have comfy cloud this also uses all the leading models and this is hosted on massive gpus and then we also have the option to render open models locally on your own system your own gpu and then everything will land and store in your computer and you can drag a render straight into your edit Don't get me wrong, Higgs Field is a very great tool.
But of course, some things they put out is questionable. Here we have Higgs Field. This is a polished all -in -one website.
It requires no setup at all. It's generally easy to use. But the downside is it's about $94 to $130 every month.
Even if you have a quiet month that you don't use it. you will still need to pay that monthly subscription. And if you cancel, you basically have no access anymore.
Also, your credits expire. They never accumulate. And even your top ups will expire as well.
So you're essentially wasting money at that point. And I can imagine a lot of people are not using their credits or they need to top up every month. And one major disadvantage is that everything stays inside of Hicksfield on their cloud.
You never really own it and they even can use your stuff. Now with AI Video Studio, it's a one -time payment, no subscriptions ever, and it lives right inside of DaVinci Resolve and connects really well to your timeline. We can pay per clip or render free and locally on your own GPU, which is something other platforms will not allow you to do.
Every render you generate, every prompt will be stored and saved on your disk. The only downside is when you want to use free local models, you need a very capable GPU. When hardware and AI will become better, my goal is to have everyone render all these ai models locally for privacy and everything this is so much better and as a comparison here is hicks field with 99 or 130 dollars per month for 3000 credits when you do a lot of volume every month this can be nice it's about one dollar per clip but we can run out of credits and then you have to buy more and more and it can be quite expensive over time for top up credits also expire after 90 days when you have a quiet month you also have to pay this same amount same subscription unlike fall for example here we just pay per generation so if i'm working on a project for example and i just need 10 or 5 clips using cdance i can just pay that per generation and then when i'm not generating for a few weeks i will not be charged for a five second clip it can range anywhere between 30 cents to one dollar if you want to use cdance 2 .5 for example with five seconds 720p it can be around two dollars so if you max everything out have 4k 30 seconds
everything to the max then of course it can be quite expensive per clip so keep that in mind when you are generating then we also have the comfy cloud mode this is a credit system this can range anywhere to 20 a month up to 100 a month same here if you want to use expensive models the costs can add up and your credits can go fast but when you top up credits it will keep its value for one year and it's also very handy you can try it out 100 for free with 400 credits then of course if you have a beefy very nice gpu you can run it locally for free zero dollars the only thing you need to pay is of course your gpu and the electricity bill we have unlimited renders and everything stays private you can generate whatever you want and that is everything you need to know about the payment system within the plugin now for the first example you want to have a transition between two shots the first one is this car drifting and the second is this pov shot in the car and we want to have an ai transition between the outside of the car and the inside of the car and we can do that 100 locally using minimax h3 for that i will go to the node studio then right click add a reference image and we can add a reference image right from our resolve timeline
And this will be our starting frame. Then we can also scrub to our end frame, which is the POV. Then we can also add in a reference frame.
Very basic start and end. Then we can also add our video model, which is going to be the Minimax H3. And for this, I will choose image to video model because we are working with reference images.
Then I will change the duration to 2 seconds because it's just a transition and the max quality for now is 720p. Then I will also add in a detailed prompt for the transition that needs to happen. And also important for this model is that you state that it needs to have no cuts, no dissolve and no cutaways.
We just want a smooth transition. you can even go to the ai assistant and ask it to create a prompt for us you can also go to cloud for example to get the right prompts then we can add this node into the prompt input of minimax and then we can hit run if you have a really good gpu the minimum is 24 gigabytes of vram But if you have higher, that's even better.
If you're just generating images, you don't need that much power. And basically how this works, if you go to the settings here, you can see we have our local models. And these are all with downloadable links that automatically sets up the nodes and everything connected to CompVUI.
So you don't have to do anything. Just make sure you have CompVUI installed. You install all the models and then you can start using it.
And here we go. Now we've made transition. like this and we can still tweak it a bit to make it better but this is already very nice.
We can also add in a duration of 4 seconds and then generate again. We don't have to worry about credits or anything like that because we are running this 100 % locally. After that it generated a variation like this.
Here it goes more on the side of the car and then creates a smooth transition to go into the POV. And I actually like the first one, so I will upscale this using the upscaler. So that way it's a bit sharper and crisper to look at.
Once you're happy with the final generation, you can either add it to the timeline or to the media pool, but we can also just drag it. onto our timeline in this example i will place it between the two clips and now we have a very smooth transition now to make it blend with our real footage i have a handy little trick for this you can offset the clip by a few frames to the right here and i will add in a smooth cut effect to make the pixels blend together that way the transition is seamless and we go from the real footage onto the ai render and the same thing for the end of the clip i will trim it and then add in a smooth cut effect And there we go, now we have the final transition.
We can even add some color grading to this to make it a bit better. And it's just crazy that this was run locally inside Resolve without ever needing to do complex nodes and all that stuff. We just have it right here, quick transition, making your edits a bit more seamless and nicer to look at.
and here's another local video generation just to test a dog jumping and minimax made this from it and it even made a cut from the dog jumping and catching the frisbee now as a second test i want to composite in a new car for that we can use the edit video function and i will use the draw to video with this we can select a frame of our video then paint on top of it just so the ai has a reference on where it needs to generate then i will add in a reference for the new car i want to swap i will upload two reference images for this and then create a prompt like this replace the car with the corvette from reference image 1 and 2 keep the same angle and use the license plate from the original car Then I also add in a frame from the original car so it has some reference for their license plate.
That's basically all you have to do. I will also say composite the car into the scene. Keep the original motion from the video.
And to make it more efficient, I will choose frame first. So we generate a reference frame before we start generating our video. That way, when we are running it on the cloud, it's much cheaper to generate images first and then videos.
So for that, I would choose GPT -2, Image Edit, and then Quality High. And make sure you use a video -to -video model for this. So I would choose Seedance 2 .5.
For now, we can only go to 720p with Seedance 2 .5. But in the future, we will also have HD and 4K. just like we have in cdance 2 .0 for the duration i will say auto and then generate our image then it generated a reference image like this and that is exactly what we wanted it keeps our original license plate and swaps it with the new blue car once you are happy with this you can double check the settings and then animate your clip then it will take a few minutes to generate the final video and here we have a final result like this the car is composited in nicely and it looks very realistic considering the time this took normally you would have to do some complex cgi work to make this this work like that but it's crazy what ai can do nowadays now i can also upscale this so the clarity of the video is a bit nicer but of course it's not going to be the same than doing a true 4k animation So for that, I also did another test using Seedance 2 .0 with a 4K model.
And as you can see, this result looks much better and more realistic because we have the max quality. The only thing that's happening here is the hand is a little bit weird. And also my face gets distorted once I get into the car.
So it's not perfect, but it's still very usable in my opinion. It even changed the original movement a little bit. Just to compensate with the height of the car.
Which is also pretty good.
Now another VFX shot we can do. Is let's say we want to have mountains. Distant mountains in the background.
And we want to remove these cars here. Then also in edit video we can describe what we want to change. Say something like remove the cars from the road.
And add large distant snowy mountains along the horizon. Then I will choose Seedance 2 .0 video to video. And for this one, I will choose 1080p.
And duration auto match clip. Then we can also hit generate. And let's see what it's going to do.
You can also check the backend. So here you're in Comfy Cloud. And we can see our video is actually running and rendering using our credits.
if you're using fall you are paying per generation for example this scene i relit this was about one dollar and one cent for this four second clip and now the video is done generating using cdance and as you can see it removed the cars perfectly and it also added in the mountains We can then overlay that on top of our original clip.
When I set the composite mode to difference, we can see exactly how it lines up. And sometimes AI tends to drift from your original clip, but this video did it really well and it's very close to the original movement. I will also add in some masks using Fusion.
That way we keep as much of our original footage and only replace the parts that are generated with AI. this is the final composite looks very clean and considering the time this only took about three minutes to generate so i mean normally for vfx artist this would have taken hours to paint out and do the compositing and everything so now we just generate place some masks and that's it Now as a bonus, another thing we can do using Minimax locally is restyling our clip.
For example, here we have a performance clip of a girl dancing in a field and we can add in that video onto the notes as well. So that way we have a preview of our video. For the restyling, I added in a prompt, then also generated some reference images of the new look.
Here we made the scene of the doll and also a second reference image of the face itself. Then we feed that in to Minimax H3 with the video images and prompt. Then it generated a final restyled video just like this.
You can also upscale this or even reframe this if you want to. And there's all sorts of things we can do with this workflow. And that was the full overview on how to do AI visual effects using AI Video Studio for DaVinci Resolve.
The plugin is now available on Gumroad as a one -time payment. So if you're interested, go check it out. We also have a community on school which teaches exactly the workflows you have seen today.
Links are in the description and see you guys on the next one. Peace.
The Hook
The bait, then the rug-pull.
Every leading AI video, image, and 3D model, running inside the DaVinci Resolve timeline itself, cloud or 100% free on your own GPU. That's the pitch for a plugin its creator says removes the need to ever leave your editor to generate a shot.
Frameworks
Named ideas worth stealing.
03:10list
Three render modes
fal.ai (pay per generation, cloud)
Comfy Cloud (credit system, hosted GPUs)
Local GPU (free, private, unlimited)
The plugin routes every generation through one of three backends, letting the user pick cost and privacy tradeoffs per shot.
Steal forany tool comparison that wants to show a cost ladder from paid-cloud to owned-hardware
CTA Breakdown
How they asked for the click.
VERBAL ASK
16:37product
“The plugin is now available on Gumroad as a one-time payment. So if you're interested, go check it out. We also have a community on Skool which teaches exactly the workflows you have seen today.”
Direct, single ask at the very end after establishing cost credibility throughout the whole video; backed by a secondary community/course upsell and links in the description rather than a second product pitch on camera.
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 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.
One editor runs nine real tasks through DaVinci Resolve's new native AI integration to find out where it saves real time and where it's just slower than doing it by hand.
Team 2 Films runs Claude through four real edit jobs, project setup, a rough assembly cut, TIFF-cached rendering, and color grading, to see what Resolve's new AI agent integration can and can't do.
00:16
00:31
00:40
00:53
01:09
01:16
01:47
02:00
02:13
02:25
02:51
03:03
03:16
03:29
03:41
03:54
04:07
04:19
04:32
04:45
04:57
05:35
05:48
06:01
06:13
06:26
06:39
06:51
07:17
07:29
07:42
07:55
08:07
08:20
08:33
08:46
08:58
09:11
09:24
09:36
09:49
10:02
10:14
10:27
10:43
10:56
11:00
11:18
11:43
11:56
12:08
12:21
12:34
12:46
12:54
13:14
13:24
13:37
14:02
14:15
14:28
14:40
14:53
15:06
15:18
15:31
15:39
16:09
16:22
16:48
17:21
05:42
17:33
18:59
24:14
14:51