Ninety minutes of good conversation is sitting on a drive, and you already know how the next three hours go: scrub, guess, trim, watch it back, guess again. Nobody sells you a tool for that part, because recording is the part that's easy to sell.
Riverside — it dropped the .fm and moved to riverside.com in May 2026 — is very good at recording. It pulls clean, locally recorded, separate tracks out of people scattered across the internet, and one guest's bad wifi doesn't wreck the take. That is a hard problem, and BitterClip's one-person browser recorder doesn't try to solve it.
What happens after the session ends is a different story. The recording lands in an editor built around transcript deletes, timeline trims, and quick auto clips, and the pricing tells you where the attention went: footage you shot somewhere else draws on its own small allowance you can't even check, downloading each guest's track is a monthly quota, and the API sits behind a custom-priced plan. Recording is the building. Editing is the annex.
BitterClip is the annex, built out. There is a recorder in it: prop your phone up, hit "Record in browser" from the project's Sources tab, or "Record here" in the editor to drop a take straight into the edit at the playhead. It takes camera and mic, or your screen in desktop Chrome, up to 1080p, and it uploads while you record instead of leaving a big file on your device, so transcription starts when you press stop. It is one person, one device, signed in — no guest links, no separate tracks, no 4K. So mostly you bring the file — a Riverside session, a Zoom call, a phone video, a camera card — and from there it's an episode you can actually work on, and find again months later by what was said, who said it, or what was on screen. It runs in a browser with nothing to install, and it'll get you through a fix from your phone.
Both products let you talk to an assistant about your video. The difference is where you're standing when you do it. Riverside's AI works inside Riverside. BitterClip's works from ChatGPT, from Claude, from any MCP client, or right inside the editor — the same moves wherever you ask. Anything you can click, you can ask for. Say "cut the tangent at 14:20 and tighten the intro" in the assistant you already had open, and what comes back is an ordinary edit. Open it, drag it, change your mind by hand, undo it.
The other difference is quieter, and you feel it every day rather than once. BitterClip cuts against the actual audio, so an edit lands on the word instead of near it — no clipped syllable, no half-swallowed first word, no nudging the edge a frame at a time until it stops sounding wrong. Every clip remembers where it came from, so one click puts you back at that spot in the full recording when someone asks where it's from. The review link you sent last week still plays the version they watched. When the first cut isn't right, you fix that cut rather than rolling the dice on a new one, and undo goes all the way back. Nothing scores your moments and tells you which ones are good. You decide, or you ask.
Keep recording where recording is best. Finish where finishing is the whole product. Bring one Riverside session across and see what the same conversation looks like done.
