[megous]

Shouldn't be hard to make opensource one. CSI capture under Linux is fairly simple with v4l2 API, SG2002 SoC has HW accelerated JPEG codec, and that's exposed as mem2mem v4l2 API too, usually. USB HID gadget emulation is fairly easy under Linux, too, via /dev/hidg# device. It's basically just shuffling some buffers around in userspace in video->net direction, and translating between browser input events and HID in the other direction.

It's probably 2-4 weekends worth of work for someone who already worked with all of this in the past to combine it together and add a basic web interface. (Assumming they will use sophgo's Linux kernel branch and will not try to mainline kernel support, first)

[zepan]

"CSI capture under Linux is fairly simple with v4l2 API" but there have no v4l2 support. I'm the maker of NanoKVM, we also want use PiKVM/openBMC, but bsp sdk don't support V4L2, most IPC SOC SDK use something like MPP API, not V4L2 API. We have to build software from scratch，it takes about 2~3month. Also other community developers have get NanoKVM samples 3month ago to port PiKVM, but still not finished(seems will finished few month later). We want make it as a serious product, not a protype tool. 80% function will finished in 20% time, but rest 20% function/bug need 80% time.

[megous]

> but bsp sdk don't support V4L2 .... use something like MPP API

Well that sucks then. I hoped it's in similar state like Rockchip, where the required things already work over V4L2 API (in both BSP and upstream).

But that just means that Linux upstreaming will have to come first. I personally don't see why anyone would want to waste time with some random vendor's media API and its quirks when it will get replaced by v4l2 in the future.

To me 20/80 also means that I need 20% of the features of a serious product and prototype is often enough, as long as it's reliable. :)