Match the symptom before changing settings
Start with browser permission, the Windows/Linux default input and the physical microphone connection.
This matches the common Godot AudioEffectRecord failure reported by the developer and many players.
Lower music, check the selected output and raise the input level without clipping.
Remove virtual routing and audio interfaces for one test. The game does not currently offer robust latency calibration.
Local browser tool
Microphone and playback checker
This tool asks the browser for one microphone, shows the reported input settings, and can make a short recording that stays in this tab. It does not send audio to ReStory Fix or test the game's Godot recording path.
No microphone access requested yet.
If this page records and plays your voice but The Choicer Voicer stays silent, your browser, OS permission and basic input are working; the remaining problem is more likely inside the game, its selected output, or the known Godot recording bug.
Safe game checks, in order
- Run the official free demo. Test recorded playback before using the paid build or a long Dub session.
- Select plain system devices. Choose the actual headphones/speakers and microphone rather than a surround, Sonar, virtual mixer or broadcast channel.
- Restart after changing devices. Recent official comments include cases where selecting a new input required a restart.
- Use one or two output channels. The developer says multichannel output is a leading trigger for the Godot record-effect bug.
- Record one short sentence. Confirm that “Hear Recording” works; do not rely on the live meter or waveform alone.
- Only then restore extras. Re-enable EQ, virtual routing, surround sound or an interface one component at a time.
What channels and sample rate can tell you
The local tool reports what the browser exposes for the selected input and its own audio context. It cannot directly read the Windows speaker-mode setting used by Godot, so a browser result is a diagnostic clue—not proof that the game sees the same channel layout.
- Channel count: one or two input channels is the simplest test path. The official developer evidence focuses more strongly on multichannel output than on the microphone itself.
- 44,100 Hz: several players reported success after choosing 44.1 kHz and one/two channels. The developer asked for more testing, so label this a community workaround rather than an official requirement.
- 48,000 Hz: a browser or interface reporting 48 kHz is not automatically wrong. Change it only as a controlled test and remember the original setting.
- Virtual audio: it can help Discord routing, but it also adds another layer that can break or delay recording.
Which 0.5.2 build should you test?
The official store currently lists 0.5.2 full files for Windows and Linux. An August 6 official devlog separately describes 0.5.2-dev2 as an experimental mic-test build made with a custom Godot export; at the time of that post, the developer explicitly said it was not a final release. Check the current official file list and devlog before switching builds, and do not download reuploads from file-sharing comments.
Frequently asked questions
Why can I hear Live Microphone but not Hear Recording?
The live monitor and the game's record effect use different paths. The developer says the Godot record effect can fail even when the microphone itself is detected.
Does a moving waveform prove the recording was saved?
No. Multiple current reports describe a moving meter or waveform followed by silent playback. Always play back a short sample.
Should I delete save.json?
Not as an early fix. A community comment reports that resetting it changed a muted configuration, but this is not an official universal solution. Back up the entire game data folder before touching any save or settings file.
Sources and verification
- Official store page — current 0.5.2 files, free demo and known microphone warning.
- Official 0.5.2-dev2 mic-test devlog — experimental status and mixed test feedback.
- Official sticky microphone discussion — developer explanation, channel hypothesis and labelled community workarounds.
- Current silent-playback report — confirms the exact “detected but silent” question remains active.