Standard WebRTC video conferencing in multi-party calls relies on Selective Forwarding Units (SFUs) that terminate DTLS-SRTP encryption at the media server, exposing video and audio in plaintext on the server. WebRTC Encoded Transform (formerly known as Insertable Streams) introduces a programmable hook directly into the WebRTC pipeline via a TransformStream. It intercepts raw encoded video (H.264, VP9, AV1) and audio (Opus) frames between the media encoder and the RTP packetizer, allowing client applications to encrypt frame payloads using custom ciphers (such as SFrame / RFC 9605) before transmission, guaranteeing zero-trust SFU privacy.

SFrame encrypts only the media payload of each encoded frame using AES-GCM or ChaCha20-Poly1305, while preserving unencrypted codec metadata headers (such as NAL unit types, keyframe indicators, temporal/spatial layer IDs, and resolution). It appends an SFrame header containing a Key ID and an 8-byte frame sequence counter. Because the SFU can still inspect frame types and scalability layers without decrypting the video content, it can perform intelligent selective forwarding and bitrate adaptation while the content remains 100% confidential between participants.

Video and audio processing is time-critical: at 60 frames per second, each frame must be processed, encrypted, and queued within less than 16.6 milliseconds. If transform operations run on the browser's main thread, garbage collection pauses, DOM rendering, or heavy JavaScript execution can cause dropped frames, audio distortion, and buffer bloat. Modern browsers allow developers to transfer readable and writable streams into a DedicatedWorker via postMessage() using transferable objects ({ transfer: [transformStream] }), achieving jitter-free real-time performance.

WebRTC Encoded Transform and WebCodecs operate on compatible data representations. RTCEncodedVideoFrame payloads can be extracted and fed directly into a WebCodecs VideoDecoder for background machine learning (e.g. gesture recognition, computer vision, or real-time super-resolution). Conversely, synthetic or transcoded VideoFrames from WebCodecs VideoEncoder can be converted into RTCEncodedVideoFrames and pushed into an active WebRTC broadcast pipeline.