One-page reference: the shared face-landmark map and how beauty, gaze, and AR effects each use it; MediaPipe Face Landmarker outputs (478 landmarks, 52 blendshapes, transformation matrix); where effects attach in a WebRTC call via Insertable Streams (MediaStreamTrackProcessor → worker → VideoTrackGenerator); the 33 ms frame budget at 30 fps and the GPU-vs-CPU choice; the EU AI Act line between altering a face and reading its emotion; and a build-vs-buy checklist (MediaPipe vs Banuba/Snap/DeepAR, Maxine vs Apple for gaze).
Download free PDF
The book · Volume 7 of 7
AI for Video: A Complete Guide to Intelligence in Video Products: From Real-Time Pipelines and Agents to Serving, Cost, Compliance, and Verticals
This cheat sheet shows how beauty, gaze and AR face effects share MediaPipe landmarks and attach to a WebRTC call within a 33 ms frame budget. Volume 7 covers this in depth: where AI fits in a WebRTC pipeline, call latency budgets, GPU serving with TensorRT, and EU AI Act limits on reading faces. Written by Nikolay Sapunov, CEO at Fora Soft.
Adding face effects to video calls without breaking the 33 ms frame budget?
Tell us your platforms and effects, and we will choose MediaPipe or an SDK and wire it into your WebRTC pipeline. 250+ video products shipped since 2005.
You get a realistic estimate, technical recommendations and next steps. Free, no obligations.
