WebRTC actual's complezity is very high. WHIP seems to be the standard path for most apps to integrate, but it does rely on an exterior service to actually do anything.
Hypothetically ffmpeg could be an ICE server for peer-connecting, do SDP for stream negotiation possibly with a side of WHEP (egress protocol) as well, could do SCTP for actual stream transfer. Such that it could sort of act as a standalone peer, rather than offload that work to a gateway service.
Worth noting that gstreamer & OBS also are WHIP based, rely on an external gateway for their WebRTC support. There's not one clear way to do a bunch of the WebRTC layer cake (albeit WHEP is fairly popular I think at this point?), so WHIP is a good way to support sending videos, without having to make a bunch of other decisions that may or may not jive with how someone wants to implement WebRTC in their system; those decisions are all in the WHIP gateway. It may be better to decouple, not try to do it all, which would require specific opinionative approaches.
Hypothetically ffmpeg could be an ICE server for peer-connecting, do SDP for stream negotiation possibly with a side of WHEP (egress protocol) as well, could do SCTP for actual stream transfer. Such that it could sort of act as a standalone peer, rather than offload that work to a gateway service.
Worth noting that gstreamer & OBS also are WHIP based, rely on an external gateway for their WebRTC support. There's not one clear way to do a bunch of the WebRTC layer cake (albeit WHEP is fairly popular I think at this point?), so WHIP is a good way to support sending videos, without having to make a bunch of other decisions that may or may not jive with how someone wants to implement WebRTC in their system; those decisions are all in the WHIP gateway. It may be better to decouple, not try to do it all, which would require specific opinionative approaches.