I was wondering if anyone here has tried implementing real-time video effects (like the ones people use on TikTok) on STM32 or other microcontrollers. Since those filters need fast image recognition and processing, I’m curious whether this can realistically be done on-device with STM32 AI tools, or if it almost always needs to offload to a stronger processor/server.
Has anyone experimented with this kind of setup? What kind of trade-offs did you run into in terms of performance, memory, and power consumption?
