wtf is tributary?

Tributary is a self hosted, open source remote recording studio for podcasts and video interviews, built in the style of the popular Riverside service. Guests join a live video call through a web browser, but instead of relying only on that call's video quality, each participant's camera and microphone are recorded locally on their own device at full quality. Those recordings are saved in small chunks and uploaded in the background as they go, so the host ends up with separate, synchronized, high quality tracks for every participant even if the live connection was choppy. Before recording starts, guests get a lobby screen to check their camera and microphone and pick a quality setting. During the session, hosts can mute guests, share their screen, use a synced teleprompter script, and set up an optional waiting room where guests wait to be let in. Recording is designed to survive crashes: every chunk is saved to the browser's local storage before it uploads, uploads retry automatically, and if a browser tab crashes mid recording, reopening the page finds and finishes the unfinished uploads. Once a session ends, an ffmpeg based pipeline assembles each guest's chunks into individual video and audio files, plus a combined mixed video and audio file with everyone properly aligned in time. The project also includes automatic transcription with speaker labels, and a built in editor that lets you cut a recording by selecting words in the transcript, since the video follows the text. Finished edits can be exported as social media clips in different aspect ratios, or as a project file for professional editing software like Premiere Pro or DaVinci Resolve. Tributary is built with Node, React, TypeScript, and ffmpeg, and also includes a command line tool that exposes its editing features so an AI coding agent can automate tasks like removing filler words or generating clips. It is released under the MIT license.