Choose the lineup
Select one to four local players and enter names. Everyone shares the same device and microphone.
Browser voice game · Play instantly
Hear a line, perform it, and face five computer judges—solo or with up to four local players.
Loading
The game needs JavaScript and a modern browser with microphone support.
What is verified
The Choicer Voicer is a browser voice-imitation party game: hear a short character performance, copy its timing and vocal movement, replay your take, and face five computer judges.
This edition supports one to four local players and runs with original TypeScript, Web Audio analysis, interface design, scoring logic, and rights-cleared dialogue performances. It does not use commercial movie audio.
How to play
Every recording stays in this tab. Use headphones so the reference clip does not leak back into the microphone.
Select one to four local players and enter names. Everyone shares the same device and microphone.
Play the original performance, read the source and character card, then imitate the dialogue line in one timed take.
Five computer judges convert the local similarity analysis into 0–5 approvals. An exceptional match can reach 6/5.
Public feature map
The browser edition focuses on a complete, replayable party loop with transparent local scoring and no account requirement.
Reference clip, local recording, computer judging, 1–4 local players, five votes, and a final leaderboard are playable above.
Each JSON scene carries a quote, work, character, scene, rights status, audio file, and scoring profile. Unlicensed movie or community media is not bundled.
Pass one device between up to four named players. Microphone recordings remain temporary objects inside the current browser tab.
Scoring guide
The game does not judge whether you have the same natural voice as the character. Instead, it measures the shape of the performance you just heard. A successful take follows the reference line’s overall length, changes in loudness, and rising or falling pitch. That makes careful listening more useful than trying to force an unfamiliar accent or vocal register.
Duration contributes 25 percent of the similarity score. Starting clearly, finishing near the reference ending, and keeping pauses in roughly the same places will help. Speaking every word correctly but rushing through the middle can lower this part of the result. Before recording, listen once for meaning and a second time for pauses, held sounds, and the final beat.
Energy contour contributes 35 percent. The browser samples how the performance becomes louder, softer, sharper, or calmer over time, then compares that curve with the reference. You do not need to shout. A line that begins quietly and ends with emphasis should keep that same dynamic journey at a comfortable microphone level.
Pitch shape contributes 40 percent. The comparison looks at how your voice moves rather than demanding the exact vocal register of the recorded character. Follow a question that rises, a dramatic fall, or a sustained note in the same direction. This lets players with different voices compete on phrasing while still making the challenge recognizably about imitation.
After those three values are calculated locally, five computer judges apply different weighting profiles. One may care most about rhythm, another about pitch, and another about energy or balance. The panel converts those views into zero to five approvals and a one-word reaction. A rare exceptional match can trigger the playful 6/5 result, while the detailed breakdown still shows what to practice next.
Performance tips
A higher score usually comes from reproducing the reference performance in layers: rhythm first, then energy, then pitch movement. These practical checks work in both solo and local party modes.
Tap the main beats with a finger while listening. Notice the first sound, the longest pause, and the final word. On your take, keep those anchors stable even if the words between them are not perfect.
For a difficult scene, use the replay control before recording. One intentional second listen is more useful than repeatedly guessing after a low score.
Instead of focusing only on the text, ask where the character becomes confident, uncertain, angry, amused, or quiet. Those emotional changes often create the energy curve that the browser can compare.
Keep the microphone about a hand’s width away and avoid moving it mid-line. A stable distance lets real vocal changes show up instead of accidental volume jumps.
Listen for rises, falls, flat stretches, and held endings. Hum the melody of the sentence once if needed, then add the words while keeping the same movement.
Do not strain for the character’s exact high or low voice. The Choicer Voicer can compare contour across different natural registers, so controlled movement is more valuable than an uncomfortable impression.
Solo mode is the quickest way to learn a scene. You perform all four selected dialogue clips, see the three-part breakdown after each round, and can retry a line before moving on. Use it when you want to understand why one version scored better than another or when you are testing a new microphone.
Group mode is designed for a shared room rather than an online lobby. Two to four named players pass one device around and perform the same scene before the game advances. Because everyone hears the identical reference, the final leaderboard compares different interpretations of the same material. Headphones help prevent the reference audio from being captured as part of a player’s take.
The game works best in a current browser that supports microphone permission and MediaRecorder. On a phone, make sure another app is not already using the microphone. On desktop, select the intended input in browser or operating-system settings before starting the sound check. The live input meter should respond to normal speech without staying pinned at maximum.
Choose a reasonably quiet room, silence notifications, and keep the device still. Built-in phone and laptop microphones are sufficient for casual play; an external microphone is optional. If permission was denied, use the site controls beside the address bar to allow microphone access, then reload the page. The recording exists only for replay and scoring during the current tab session.
Guides
Pick a guide.
Control starts, pauses, and pace.
Try nine low-prep formats.
Build a choice in layers.
Prepare gently before a scene.
Carry speech without shouting.
Control one variable at a time.
Keep expressive lines clear.
Build a repeatable voice recipe.
Plan phrases and recovery breaths.
Return gently to speaking voice.
Mark listener, beats, and turns.
Practice six original ad reads.
Test distance and angle.
Review a take with evidence.
Respond to new information.
Practice first takes.
Make thoughts playable.
Control one variable.
Plan work and recovery.
Study accents without caricature.
FAQ
Review audio, privacy, scoring, and support.
Yes. The playable experience at the top of this page runs directly in a modern browser and includes solo practice plus local party play for up to four players.
The library contains original film-style performances and public-domain literary dialogue rendered as purpose-made voice assets. It does not bundle commercial movie audio.
It remains a temporary browser object in this tab. There is no upload endpoint, account, or server-side voice storage.
The browser scoring model compares timing, energy contour, and pitch movement, then maps the result to five judge votes.
It is designed for current mobile browsers with microphone and MediaRecorder support. HTTPS is required after deployment.
Not in this build. Pack files contain third-party media and should only be used when their author grants permission.