DeepMake
DeepMake brings open source generative AI into Adobe After Effects, running entirely on your own GPU. Generate images from prompts, mask faces and objects automatically, and upscale footage up to 4x, with no usage limits.
What is DeepMake?
DeepMake is a layer of open source generative AI bolted onto professional post-production software, running on the machine in front of you rather than in a rented data centre. Its homepage promise is unlimited creativity with generative AI on your computer, and the architecture follows from it: the tool is explicitly not a cloud rendering service, it drives your own GPU inside a workflow you already use.
Today that workflow means Adobe After Effects, where DeepMake installs as a plugin and appears under Effect, then DeepMake, then AI Plugin Renderer. Support for Nuke and DaVinci Resolve is announced but still listed as coming soon. The system splits into host plugins, which connect DeepMake to the editing application, and effect plugins, which carry the models themselves.
Five capabilities ship today. Text-to-image and image-to-image generation both run on Stable Diffusion. Smart face masking uses Bisenet and lets you choose which parts of a face to include. Smart object masking combines SegmentAnything with GroundingDINO, so a character, a background or a prop can be isolated simply by typing what you want. Image and video upscaling relies on BasicSR and reaches 4x, sharpening and denoising as it goes; it is also sold as a standalone upscaler outside the plugin.
None of these models is locked in. A Models Manager lets you pick a different model from a vetted list, or download one yourself from an open source repository and point DeepMake at it. New AI tools are promised on a monthly cadence.
The company behind it, Arkavian Inc trading as DeepMake, describes itself as an open core business: the code sits on GitHub, the employee handbook is public, and a published ethical manifesto sets explicit limits on use. Its stated lineage is FaceSwap, the open source deepfake project active since 2017, with DeepMake itself founded in 2022. A wider ambition is visible on the site, a standalone AI-native editing and post-production suite in both desktop and cloud versions, whose first component is AI rotoscoping. That part is still a beta signup rather than a product.
What it does
- Generate images from a text prompt directly inside After Effects
- Refine, augment or transform an existing image with image-to-image
- Mask objects, characters and backgrounds by describing them in plain text
- Mask faces automatically, choosing which facial features to include
- Upscale images and video up to 4x while sharpening and removing noise
- Swap the underlying open source model through the Models Manager
- Run every operation locally on your own GPU, without usage caps or per-render fees
When to use DeepMake / When not to
A quick filter to help you decide if DeepMake is the right fit.
When to use DeepMake
- Motion designers and VFX artists who already work daily inside Adobe After Effects
- Studios and freelancers with a capable local GPU who want to avoid cloud credits and per-render fees
- Editors facing repetitive rotoscoping and masking work on faces, objects or backgrounds
- Teams that prefer to keep client footage on their own machines rather than upload it
- Open source enthusiasts who want to swap in their own Stable Diffusion models
When not to use DeepMake
- Anyone working on a machine without a dedicated GPU, or on an Intel-based Mac
- Editors living in Premiere Pro, Final Cut Pro, Nuke or DaVinci Resolve, none of which are integrated yet
- Teams wanting a browser or mobile app, since DeepMake is desktop software only
- Buyers who need a published price before they are willing to talk to a salesperson
- Organisations whose procurement requires a DPA, security certifications or a written GDPR commitment
How to use DeepMake
A typical end-to-end flow, from setup to results.
- Check your hardware first: an Nvidia 20xx or later GPU with 8 GB of VRAM on PC, or an Apple Silicon Mac with 16 GB of RAM or more
- Create an account, sign in, and open the Account page to reach the downloads
- Choose Windows or macOS, run the installer, and restart the machine if prompted
- Alternatively install manually: download the repository zip from GitHub, install Anaconda, then run conda env create -f environment.yml
- Open After Effects and select the layer or element you want to work on
- Choose Effect, then DeepMake, then AI Plugin Renderer to load the panel
- Find the panel in Effect Controls, pressing F3 if the window is not visible
- Click Login in the DeepMake panel and enter your email address and password
- Open the Plugin Manager with the Manage button and install the plugins your plan includes
- Restart After Effects so the newly installed plugins become usable
Pros & Cons
Pros
- Everything runs locally: no usage caps, no per-render billing and no footage leaving your machine
- The Community tier is genuinely free and advertised with unlimited usage
- Models are open source and interchangeable, including models you supply yourself
- It slots into an existing After Effects workflow instead of demanding a new tool
- Open core transparency: public GitHub repository, public company handbook and a published ethical manifesto
- Ten video tutorials cover installation on both platforms and every feature
- The upscaler is also available as a standalone application
Cons
- Demanding hardware: a dedicated 8 GB Nvidia GPU, or an Apple Silicon Mac with 16 GB of RAM
- Only one host application is really integrated; Nuke and DaVinci Resolve remain coming soon
- No public price for the paid tier: Studio is listed as Custom and routes to a contact form
- A Creator tier is referenced throughout the support documentation but no longer appears on the pricing page
- No terms and conditions are published anywhere on the site
- No DPA, no certifications and no mention of the GDPR, which will stall many procurement processes
- Slow public cadence: no blog post since March 2024 and an empty careers page
Pricing & Plans
A permanently free plan is available. The Community tier is listed at no cost with unlimited usage, and it is the only tier for which a price is published. The paid Studio tier is shown as Custom and requires contacting the company, so no entry price and no currency can be quoted. The pricing page offers a monthly and yearly billing toggle, with yearly billing advertised as including two months free, but no amount is attached to either option. No free trial and no refund policy are mentioned.
- After Effects integration
- text-to-image
- image-to-image
- latest open source model updates
- video and image upscaling
- smart face masking
- smart object masking with text prompts
- and access to new AI features
- everything in the Creator Plan
- custom integrations
- multi-computer rendering
- white glove deployment
- unlimited usage
- dedicated email support
- plus custom development
- training and features
- referenced in the support documentation
- where BasicSR
- Bisenet and Gsam are listed as Creator tier and email support is restricted to Creator and Studio
- but no card and no price appear on the pricing page
Data, GDPR & hosting
A consolidated view of how DeepMake handles your data.
GDPR overview
The site never uses the words GDPR or General Data Protection Regulation, and makes no compliance claim in either direction. The privacy policy does list the six data subject rights verbatim: access, rectification, erasure, restriction of processing, objection and portability. It commits to a one-month response time and notes that a small fee may be charged for an access request. Requests go to hello@deepmake.com. No Article 27 EU representative is designated, no data protection officer is named, no Data Processing Agreement is published or offered, and no certification such as SOC 2 or ISO 27001 is mentioned. The policy also carries two contradictory update dates. The rights are present in substance; the formal compliance apparatus a European buyer looks for is not.
Who owns the data?
DeepMake publishes no terms and conditions: requests to /terms, /tos, /legal and /eula all return 404, and no such page appears in the sitemap. Ownership of the images and videos you produce is therefore never addressed in writing, in either direction. The privacy policy covers only account and diagnostic data and contains no clause claiming rights over user content. In practice the company states it is not a cloud rendering service and that rendering happens on your own GPU, so footage need not leave your machine at all. The point to note here is the absence of a contract rather than an unfavourable one.
Reuse rights
Because no terms of service are published, nothing on the site restricts what you may do with the images and videos you generate, and nothing grants you an explicit licence either. The privacy policy addresses only the data DeepMake itself collects: name, email address, business name, any messages you send, crash reports and anonymised performance data. It lists the purposes for that data, which are order handling, mailing, bug diagnosis, performance troubleshooting and usage monitoring, and notes that order data may be sent to credit reference agencies to prevent fraudulent purchases. Reuse of your own output is never mentioned. Commercial users should request written confirmation before relying on it.
Data retention & training
Hosting summary
DeepMake does not disclose where user data is hosted. Neither the privacy policy nor any other page names a country, a region or a jurisdiction for storage. What the policy does name is two partner services: Hubspot, which holds contact information for the mailing list, and Sentry, which holds crash logs and application telemetry. Both are third-party processors and both are stated plainly, which is more than many comparable sites offer, but their locations are not given. The picture is different for the creative work itself: DeepMake describes itself as not being a cloud rendering service and runs its models on the user's own GPU, so images and video need not leave the workstation at all. That local-first design is the site's central selling point rather than an incidental detail. For completeness, the marketing site is served by Netlify and resolves to an address in Frankfurt, Germany. That describes delivery of the web pages, not storage of application data, and should not be read as a hosting commitment.
Things to keep in mind
Risks and trade-offs to weigh before adopting DeepMake.
- The technology descends from deepfake research, and the company's own ethical manifesto forbids using it on people without their consent
- Realistic face and object manipulation makes misleading imagery easy to produce and hard to detect afterwards
- No terms and conditions are published, so nothing contractual governs what either party may do
- No GDPR commitment, no DPA and no named EU representative, a real obstacle for European organisations
- Order data may be passed to credit reference agencies, which the privacy policy states without elaborating
- Leaning on generative masking and upscaling can erode the manual craft of rotoscoping and compositing over time
- Models can be downloaded from open source repositories, so a careless choice may import licensing or safety problems along with the weights
Setup & Integrations
Technical difficulty
Moderate. The supported path is straightforward: sign in, download the Windows or macOS installer from the Account page, run it, restart if asked, then load the plugin in After Effects and install what you need through the Plugin Manager. Ten video tutorials cover both platforms. The manual route is harder, requiring Anaconda, the command line and a conda environment per plugin. The real barrier is hardware rather than software: a dedicated 8 GB Nvidia GPU or an Apple Silicon Mac with 16 GB of RAM. Familiarity with After Effects is assumed throughout.
Deployment
Integrations
Behind DeepMake
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Does DeepMake run in the cloud?
Which editing software does DeepMake support?
What hardware do I need to run it?
Is there a free version?
How much does the paid plan cost?
Which AI models does DeepMake use?
How do I get support?
What data does DeepMake collect?
How long are crash logs kept?
Are there terms and conditions?
Should you pick DeepMake?
DeepMake makes an unusual bet: instead of selling cloud credits, it hands you open source models and lets your own GPU do the work. For a motion designer who already owns a capable machine and lives inside After Effects, that bet pays off immediately. There are no render queues, no per-generation billing, no footage uploaded to a third party, and the free Community tier is not a trial but a permanent plan advertised with unlimited usage.
The limits are just as clear. One host application is genuinely integrated; Nuke and DaVinci Resolve have been coming soon for some time. The paid tier carries no published price, and a Creator plan referenced throughout the support documentation has quietly disappeared from the pricing page, which makes the commercial offer hard to read. More seriously for anyone buying on behalf of a company, the site publishes no terms and conditions at all, never mentions the GDPR, offers no data processing agreement and names no certifications. The privacy policy that does exist is short and candid about what it collects, but carries two contradictory update dates.
Signals of pace are mixed. The product installs and works, the tutorials are thorough, and an AI-native editing suite with rotoscoping at its centre is in beta. The blog, however, has been quiet since March 2024 and the careers page is empty.
The verdict follows that split. As a free addition to an individual or small-studio After Effects workflow, DeepMake is worth installing this afternoon, and the ethical manifesto attached to a tool descended from deepfake research deserves a careful read. As a procurement item for an organisation with a compliance function, it is not ready, and the missing contractual layer is exactly why.
- Choosing a selection results in a full page refresh.
- Opens in a new window.