The problem here is that, if you prize security, privacy, or decentralization above anything else, you must make trade offs by increasing some friction.
The more centralized information you have, the better your UX can be.
On the other hand, if you prize data collection or centralization, you must make some trade offs, including increasing some friction, as well.
The less you seek to control your users, the better their experience can be.
A tiny anecdote, with Jitsi Meet you can send a link to a meeting, someone who’s never used it before can click it, and they’re there. No need to set up or be logged into any account, install an application, or even find a small web client link at the bottom of what’s more or less a malware ad.
If you primarily care about user experience and lack of friction, accepting both centralized and decentralized approaches as means to an end, you usually end up with something centralized.
Trust me, as a screen reader user who often transmits high-quality audio, I know that a web client just won't cut it in more advanced scenarios. WebRTC implementations work... kind of. Sometimes. An actual app gives you much more flexibility and control, and makes the UX much better.
The problem here is that, if you prize security, privacy, or decentralization above anything else, you must make trade offs by increasing some friction.
The more centralized information you have, the better your UX can be.