Years of testing, now public
TargetDisplay 3.0 is the first public release of the capture viewer that was apparently not content to remain a simple window forever.
TargetDisplay · macOS USB capture monitor
TargetDisplay is a native Mac viewer for USB capture devices: low-latency Metal rendering, runtime Metal display effects, exact resolution and FPS selection, audio preview, screenshots, recording, OCR, PiP, Shortcuts, MCP automation, and TargetDongle input control.
It mostly tries to be invisible: show the target, get out of the way, and stop wrapping your video feed in chrome like a hostage note.
A sharp little window into whatever your capture card is eating.
Plug in a USB capture device, pick the exact mode you want, and use the target system in a clean Mac window. Game consoles, test benches, tiny PCs, weird HDMI gadgets—bring your own questionable hardware.
Under the hood it is doing far more than a preview window should have any business doing. We regret nothing, except maybe the amount of menu code.
TargetDisplay 3.0 is the first public release of the capture viewer that was apparently not content to remain a simple window forever.
The new engine runs configurable Metal display effects at runtime, including CRT masks, scanlines, barrel warp, phosphor glow, Old TV, Pixelate, Dithered, and importable .tdfx shader packs.
Use a natural-language prompt to watch text on the target and fire Shortcuts, REST endpoints, or MCP actions when it appears. The Apple Foundation Model runs on your Mac, not in somebody else’s mystery bucket.
CRT scanlines, Old TV, phosphor glow, and enough knobs to make a normal person leave.
TargetDisplay can run full-screen Metal post-process effects on the live capture view. The effect system is runtime compiled, parameterized, preset-friendly, and built around single-file .tdfx packages you can import, export, edit, and share.
Effects are display-only: screenshots and .mov recordings stay clean. The target sees reality. You get the haunted television version, because you asked nicely.
CRT, Phosphor Terminal, Chromatic Aberration, Black & White TV, Game Boy, Dot Matrix, Vector Phosphor, and ASCII Art ship in the picker. Finally, your expensive capture card can cosplay as questionable hardware.
Old TV, Pixelate, and Dithered are adapted from single-pass ShaderToy shaders, with attribution and original licensing carried along instead of buried in a basement.
.tdfxAn effect is one JSON file: metadata, Metal source, sliders, and presets. Drop it in the Effects folder or use Import Effect. Congratulations, you are now one text editor away from GPU crimes.
Native rendering with frame coalescing, MetalFX options, sharpening, color-aware output, and the display-effect post-process pass. The pixels arrive, the pixels get shown, then optionally get weird.
MetalFX enhanced upscaling, image sharpening, metadata-aware color management, sRGB / Rec. 709 / Display P3 output, and HDR-tagged signal handling. Wild concept: video should look like video.
Select the capture device, resolution, FPS, color behavior, and audio path. It restores devices by stable identifier after relaunch or reconnect, because USB enjoys being dramatic.
Grab frames, record video, crop to the interesting bit, and run local Apple Vision OCR when the target is full of tiny text and poor decisions.
Keep the capture feed floating while you work elsewhere. Useful for installs, firmware updates, and other activities best described as “waiting, but anxiously.”
Automate status checks, frame capture, OCR, timed text logging, timed video capture, device selection, diagnostics, and emergency input release through macOS Shortcuts.
Ask the on-device Apple Foundation Model to watch OCR text from the target and trigger Shortcuts, REST calls, or MCP events when the thing finally appears. No cloud required, no “trust me bro” webhook relay.
The direct-download build includes a local MCP bridge so AI clients can observe the capture feed, OCR it, wait for changes, and—if you explicitly allow it—control the target.
With TargetDongle, TargetDisplay can forward keyboard and mouse input to the captured machine over BLE as plain USB HID. No network séance required.
Capture, effects, audio, TargetDongle, automation, diagnostics. Basically the cockpit, but with fewer unlabeled switches.
Pick exact source modes, tune the shader stack, choose audio monitoring, pair input hardware, grant automation access, and copy diagnostics without spelunking through menus named by committee.
The app keeps read access and TargetDongle control separately gated for AI/MCP clients. Observing a capture feed and driving a target computer are not the same permission, because that would be stupid in the loud way.