Dub any scene,
in a browser tab.
Transparency: dubcircus is our own product, built end to end by co-founder Evert. The numbers below come from the live site and its database.

A $5 download, one PC, one mic.
The Choicer Voicer is a voice-acting party game that took off on Twitch and YouTube: a short clip plays with the original voices muted, you read the line, and at the end your version plays back over the picture. It is an alpha sold on itch.io for Windows and Linux, and its scenes come from a community that cuts clips into packs and shares them on GameBanana.
Which means every step has friction. Download the game. Find a pack. Download that too and unzip it into the right folder. Want to play with friends? Everyone huddles around one PC and one microphone. Want to dub a clip nobody has packed yet? Learn the pack format and build it by hand.
The idea is a great one; the wrapper around it keeps most people out. We wanted the game to work the way the web works: open a link, hear the line, say it, watch it land. No install, no account, and no waiting for someone else to cut the scene.
Cut, record, premiere.
dubcircus is a free online dubbing game that runs entirely in the browser. Pick one of a thousand scenes, or paste a YouTube link and get one cut for you in about a minute; the premiere plays your voice back over the video.
The dubcircus pipeline runs in four steps. First, a YouTube link or an uploaded clip becomes a scene. Second, an AI pass separates the voices from the music. Third, the dialogue is cut into timed lines, each assigned to a speaker. Fourth, the player hears each line, records it in their own voice, and the premiere plays the dub back in sync over the video.
Cut. Paste a YouTube link or upload a clip. An AI pass separates the voices from the music bed, and the dialogue is split into lines with timings and speakers, the same structure the community used to build by hand. A scene is ready in about a minute and stays private until its creator publishes it. Hundreds of the existing community packs were converted and loaded too, so the scenes people know from streams were on the site from day one.
Record. The recording booth walks you through the lines in order, plays the original before each take, and records your voice in the browser with MediaRecorder, with the Web Audio API doing the level meter and the mix. Every take is saved the moment you finish it, so a closed tab costs nothing, and you don't hear your own performance until the premiere, which is where most of the laughs come from.
Play together. Open a live room and share a four-letter code. Up to eight people join from their own devices and their own microphones, no signup. Everyone performs every line at once, nobody hears anybody until the premiere, and the room votes on who stole the scene. Rooms run on realtime presence and broadcast, and stay open for seven days.
Stream. Streamer mode connects to a Twitch channel read-only, without asking for any permissions. Chat scores every dub out of five by typing a number, the night keeps its own scoreboard, and chat can vote on which scene comes next. The whole thing ships in eight languages, from Dutch to Hindi to Brazilian Portuguese, with an MP4 export for TikTok, YouTube or Instagram and a share link whose preview is the performance itself.
From a download and a folder, to a link.
Everything that used to be a chore is gone. Nothing to install, no pack to hunt down and unzip, no account before the first take. A first-time visitor is recording within a minute of landing on the site, and a scene that did not exist yet costs a YouTube link and a minute of waiting rather than an afternoon in a pack editor.
Playing together stopped meaning sharing a chair. Rooms put every player on their own device and their own microphone, wherever they are, and the premiere-then-vote loop turns a solo gimmick into a party format. Streamer mode does the same for Twitch, with chat as the jury.
And it was built fast. dubcircus went from first commit to a public site in under two weeks, built by one engineer inside the same pipeline we use for client work, and kept shipping: more than a thousand scenes, roughly 27,000 timed lines, live rooms, streamer mode and eight languages.
Want one of these?
Tell us where you're starting from and we'll point you at the right way in: a deep-dive build, a focused feature, or a quick production-readiness pass.
