# Aimmy: an ONNX and DirectML aim aligner for gamers with impairments > Aimmy is a Windows-only, source-available aim alignment tool built on YOLOv8, ONNX and DirectML. It is aimed at players who cannot aim reliably, and it is explicitly not open source. **Babyhamsta/Aimmy** — Universal Second Eye for Gamers with Impairments (Universal AI Aim Aligner (AI Aimbot) - ONNX/YOLOv8 - C#) - Repository: https://github.com/Babyhamsta/Aimmy - Stars: 1,601 · Forks: 704 - Language: C# - License: NOASSERTION - Published: 2026-09-10 · Updated: 2026-09-10 · Language: en - Canonical page: https://hysenlabs.com/projects/babyhamsta-aimmy ## The problem Aimmy addresses, and who it is built for Aimmy starts from a specific claim: some players are at a structural disadvantage in aiming, and the project exists to reduce that gap. The README lists the intended audience directly, including players who are physically or mentally challenged, players with untreated or untreatable visual impairments, players without a second human interface device for pointer control, and players with poor hand and eye coordination. It also names less clinical cases, such as players in hot environments whose hands become greasy during long sessions. That framing matters because it sets the design target. Aimmy is not a training tool that reports your accuracy back to you, and it is not a statistics overlay. It is a live pointer-adjustment program: you press a trigger binding, and the cursor moves toward a detected target. The README notes that it can also be used to train how you aim, but the primary loop is assistive rather than analytical. The second half of the audience is configuration authors. The README describes a model and configuration store with repository support, and a hotswap system that lets models and configurations change without restarting the application. That only makes sense if people other than the end user are producing and sharing those files, which the README confirms by pointing to community-made models distributed through the store and the project's Discord. ## Screen capture, YOLOv8 inference and cursor adjustment The README includes a Mermaid flowchart that describes the pipeline. The flow starts at the playing game system, moves to the triggering functionality, then to screen grabbing, then to YOLOv8 recognition running on DirectML and ONNX, then to a decision step. Two inputs feed that decision: X and Y adjustments, and the field of view. The output is a change to the mouse cursor position. In prose, the README states that when you press the trigger binding, Aimmy captures the screen and runs the image through AI recognition on your own hardware. The detection result is combined with your X and Y axis adjustments and your current FOV, and that combination produces the cursor movement. Nothing in the README suggests the game is queried for entity positions; the input is the rendered screen, which is why the project describes itself as universal across games. DirectML is the part that distinguishes Aimmy from aligners built on TensorRT. The README argues that DirectML plus ONNX gives high compatibility on both AMD and NVIDIA GPUs, and that AMD cards in particular would not perform well under a TensorRT-based approach. That is a vendor-neutrality argument, not a raw-speed argument, and the README does not publish latency figures to support the speed half of its claim. The detection model is YOLOv8 in ONNX format. Aimmy ships with a pre-trained model, and the models directory in the repository plus the in-app store are where additional game-specific models come from. ## Installing Aimmy and running a first session Aimmy is a Windows desktop application and the README does not describe a package manager or a build-from-source path for end users. The setup section lists two prerequisites, then a download and extract step. Install the x64 .NET Runtime 8.0 and the x64 Visual C++ Redistributable first. The README links to the runtime download page and to the vc_redist.x64.exe installer. Both are x64; the README does not mention an ARM or 32-bit variant. ```bash # Download and install these two prerequisites on Windows x64: # 1. .NET Runtime 8.0 (Desktop, x64) # https://dotnet.microsoft.com/en-us/download/dotnet/thank-you/runtime-desktop-8.0.2-windows-x64-installer # 2. Visual C++ Redistributable (x64) # https://aka.ms/vs/17/release/vc_redist.x64.exe ``` Next, get the application itself. The README is explicit that you must pick the Aimmy zip and not the Source zip from the releases page. ```bash # From https://github.com/Babyhamsta/Aimmy/releases/latest # Download Aimmy.zip (NOT Source.zip), then extract it. ``` After extraction, run Aimmy.exe. The README's closing instruction is to choose your model. That is the whole first-run flow as documented: no configuration file to write by hand, no command-line flags, and no environment variables are mentioned anywhere in the setup section. Once the window is open, the README points to the UI for everything else. The settings that change behaviour are the AI confidence threshold, the FOV, the X and Y adjustments, the trigger binding, and the choice among five mouse movement methods. The README describes these as auto-updating: changing a value changes how Aimmy aims without a restart, and models and configurations can be hotswapped the same way. ## The licence boundary: source-available, not open source This is the part most likely to be misread, and the README pre-empts the confusion in unusually direct language. Aimmy is described as a source-available product and explicitly not open source. The stated reason is that the maintainers discourage other developers from making commercial forks. The repository carries a LICENSE file and a separate SourceAvailable.md, and the GitHub licence field for the project reports NOASSERTION, which means the platform could not map the licence text to a recognised identifier. That is a practical signal: if your organisation runs automated licence checks, Aimmy will come back as unknown rather than as MIT, Apache-2.0 or GPL. The README, not this article, is where the terms are defined, and SourceAvailable.md is the file to read before you assume any right to redistribute or fork. The project's own cost model is tied to that licence stance. The README argues Aimmy needs no upkeep because it does not read game-specific data, so it would keep working even if the team stopped maintaining it. It states there are no ads, no key system and no paywalled features, and that the tool is not and will never be for sale to end users. Whether a source-available licence is acceptable is a decision for you and, in a corporate setting, for whoever reviews third-party software. ## Where Aimmy is the wrong tool The most obvious boundary is platform. Everything in the setup section is Windows x64: a Windows desktop runtime, a Windows redistributable, and an .exe. The repository is a C# solution (Aimmy2.sln) built on .NET 8 and WPF, which is a Windows-only UI framework. There is no documented Linux, macOS, console or mobile path. Questions about running Aimmy on a console or a controller are not answered by anything in the README. The second boundary is what Aimmy actually is. It moves your mouse cursor. If your goal is to review your matches, measure your reaction time, or train against a controlled scenario, Aimmy offers none of that; the README mentions training only in passing, as something you can do with it. It is a real-time assist, not an analysis tool. The third boundary is the one the README does not address at all: anti-cheat. The README is silent on which games permit this kind of tool, and the repository does not document compatibility with any anti-cheat system. The project's own search traffic shows people asking whether it is detectable or undetected; the documentation provides no answer, and neither can this article. If a ban would matter to you, that silence is the relevant fact, not a reassurance. Finally, model quality is not uniform. The README says the bundled model was trained on thousands of images and that community models vary by game and by image count. A model that performs well in one title can perform poorly in another, and Aimmy has no way to tell you which is which before you try it. ## How Aimmy differs from TensorRT-based aligners The comparison the README draws is against AI-based aim aligners that use TensorRT. The stated difference is the inference backend, and it has a concrete consequence for hardware. TensorRT is NVIDIA-specific; DirectML is a Windows API that runs across GPU vendors. Aimmy's argument is that this makes it work well on AMD GPUs, which the README says would not perform well on TensorRT-based aligners. That is a compatibility trade rather than a pure performance win, and the README's framing is worth reading carefully. It claims higher accuracy and faster performance than other aligners, with the AMD case called out as the clearest example. No benchmark, frame time or latency number appears in the README to back the general claim, so treat the cross-vendor compatibility as the documented advantage and the speed claim as the project's own assertion. A second difference is distribution. Aimmy ships as a compiled Windows application with a GUI, a model store and a configuration store. An aligner distributed as a library or a script expects you to write the capture, inference and input-injection code yourself. Aimmy's README explicitly targets users who want to tailor an aligner for a specific game without coding, and the UI is the mechanism for that. ## Maintenance, upgrades and what the repository shows The repository is not archived, and the last push was on 2026-08-20. The most recent release is v2.5.0 from 2026-01-17, preceded by v2.4.1 in June 2025 and v2.4.0 two days before that. The gap between the January 2026 release and the August 2026 push means the branch has seen activity since the last tagged build, but the release cadence itself is not fast. Upgrading is manual. There is no updater described in the README and no package feed, so moving from one release to the next means downloading a new Aimmy.zip and extracting it. Whether your models and configs survive that depends on where they live relative to the extracted folder, and the README does not document an upgrade procedure or a rollback path. That is a real gap: if a new release regresses detection on your game, the README gives you no documented way back other than keeping the previous extracted folder. The default branch is Aimmy-V2 and the solution file is Aimmy2.sln, so the current line of development is the V2 codebase. The repository also contains configs/, models/, ModelUpload.md and verification.md, which correspond to the configuration store, the bundled models, the model submission process and whatever verification the project applies to submitted models. The README does not describe what verification.md contains. On licence cost: there is no purchase price, but the terms are not those of a standard open source licence. If your workflow assumes the right to fork and redistribute commercially, read SourceAvailable.md first. ## Conclusion Aimmy suits Windows players with a physical, visual or coordination impairment who want a screen-reading aim assist without coding, and who accept that it ships as a .zip from GitHub Releases rather than through a package manager. It does not suit anyone who needs a permissive open source licence, a Linux or macOS build, or a guarantee about anti-cheat outcomes; the README and SourceAvailable.md define the licence boundary, and the repository does not document anti-cheat compatibility. Before adopting it, check that the .NET 8 desktop runtime and the x64 Visual C++ Redistributable are installed, confirm the release page offers the Aimmy zip rather than the Source zip, and read SourceAvailable.md to see what the licence does and does not permit. ## FAQ ### Is Aimmy open source? No. The README states that Aimmy is considered a source-available product and explicitly not open source, and that the maintainers discourage commercial forks. The repository includes a LICENSE file and a SourceAvailable.md that define the actual terms. ### How do I install Aimmy? Install the x64 .NET Runtime 8.0 and the x64 Visual C++ Redistributable, then download the Aimmy zip from the releases page, extract it and run Aimmy.exe. The README warns that you must take the Aimmy zip and not the Source zip. ### What is Aimmy? Aimmy is a Windows application that captures the screen when you press a trigger binding, runs the image through YOLOv8 detection on DirectML and ONNX, and adjusts your mouse cursor based on the result combined with your X and Y adjustments and FOV. The README describes it as a universal AI-based aim alignment mechanism for players who have difficulty aiming. ### Is Aimmy safe? The documentation does not make a safety claim in either direction. It describes a free tool with no ads, no key system and no paywalled features, distributed as a zip from the project's GitHub releases page, and it is silent on anti-cheat compatibility. ### Can I use Aimmy on a console or with a controller? Nothing in the README documents a console or controller path. The setup section covers only Windows x64 prerequisites, and the application is built on .NET 8 and WPF, a Windows-only UI framework. ## Sources - [Babyhamsta/Aimmy on GitHub](https://github.com/Babyhamsta/Aimmy) - [Issues](https://github.com/Babyhamsta/Aimmy/issues) - [README](https://github.com/Babyhamsta/Aimmy/blob/Aimmy-V2/README.md) - [Releases](https://github.com/Babyhamsta/Aimmy/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/babyhamsta-aimmy