// the find
videokit-ai/videokit
Low-code, cross-platform media SDK for Unity Engine.
VideoKit is a Unity SDK that bundles camera/mic streaming, video recording in several formats, background removal, native share-sheet integration, and some speech/LLM features (captioning, text-to-struct parsing) behind one access key. It's aimed at Unity developers building UGC-style apps (TikTok clones, AR filters, in-app recording) who want to skip stitching together five separate native plugins.
Real native implementations across platforms — Android AAR, iOS/visionOS xcframework, Windows DLL, WebGL static lib — rather than a thin cross-platform shim. It has an actual runtime test suite (camera/audio device enumeration, recording, human texture) under Assets/Tests, which is rare for Unity asset-store-style packages. Sample scenes exist for most major features (replay buffer, slow-motion capture, multi-camera streaming, trim video), so you can see working code instead of just API docs.
Every feature requires an access key tied to a videokit.ai account, so this is a SaaS-gated SDK, not a self-contained library — if their backend or auth service has an outage, your app's recording features degrade with it. 'Source available' is doing a lot of work here: the repo ships precompiled native binaries (VideoKit.aar, VideoKit.dll, the xcframework), so the actual capture/codec engine is closed, and only the C# glue is inspectable. The text-to-struct parsing and audio captioning features are LLM-backed calls bolted onto a media SDK, which is scope creep that adds latency, cost, and an external dependency unrelated to the core recording/streaming job.