Voice chat moderation for Minecraft.
Chat filters never see voice. VoiceSniffer does: it transcribes speech on Simple Voice Chat or Plasmo Voice, matches your rules, then mutes, warns, or alerts staff while that player is still talking. Audio stays in memory. In development, not for sale yet.
What it does
Voice is the blind spot. A slur in chat gets caught. The same sentence on a microphone does not. VoiceSniffer sits on the voice packets, not the chat thread. Staff see who spoke, which rule matched, and the sentence — during the incident, not in a report the next morning. If the processor is down, voice chat keeps working.
-
1
Listen
Simple Voice Chat or Plasmo Voice packets are held in memory. Nothing is written to disk.
-
2
Match
The transcript is checked against the rule pack for that language. 23 packs ship. English and Czech are the ones we have measured; treat the rest as alert-only until we do.
-
3
Act
Warn, mute, or staff alert — per rule. A flag is a claim with a transcript, not a verdict. Your staff decide.
Against the usual plugin
There is one serious one-time plugin in this category. These are listing and architecture facts, not a lab race. We have not published a head-to-head latency number.
| Typical listing | VoiceSniffer | |
|---|---|---|
| Price | About $20 one-time | $14.99 launch week, then $19.99 |
| Voice | Simple Voice Chat | Simple Voice Chat and Plasmo Voice |
| Processor | One worker, small Whisper model. Bursts wait in a queue. | Docker on a normal CPU. Workers sized to the machine. No GPU. |
| Audio | Public processors can write transcripts to disk | Never stored. Flagged sentence text only, 30 days. |
| Panel | None | Cloud later. Self-host is config.yml, commands, and Discord webhooks. |
What it has
One jar. Speech is checked against your rules. Staff get the player, the rule, and the sentence. Voice chat keeps working if the processor is down.
-
Simple Voice Chat and Plasmo Voice
Works with either. Players need the same mod on their client.
-
Warn, mute, or staff alert
Per rule. A flag is a claim with a transcript, not a verdict.
-
Audio never stored
Processed in memory and discarded. Flagged sentence text is kept 30 days. No recordings, no evidence clips yet.
-
Discord webhooks
Staff alerts go to your server. Self-host is config.yml and in-game commands.
-
Docker on a normal CPU
One container. No GPU. Built-in English-only mode exists; it is not for real moderation.
-
Cloud later
Pair with /vs connect after the self-host launch. The hosted panel is cloud-only.
Own the plugin. Cloud comes later.
Not for sale yet. Launch week $14.99, then $19.99. Discord gets the listing first.
Self-host · one-time
- Your CPU · 1 network · SVC and Plasmo
- All 23 rule packs · mute, warn, or alert
- config.yml and in-game · no hosted panel
Cloud · later
We run the processor. Pair with /vs connect. Four paid months unlock permanent self-host.
$5 / mo
- 50 voice users
- English
- 1 network
$10 / mo
- 150 voice users
- English + EU pack
- 1 network
$20 / mo
- 500 voice users
- Auto-detect
- 1 network
$40 / mo
- 1,500 voice users
- All packs
- 5 networks
Caps are concurrent speakers. Audio is never stored. Hosted dashboard is cloud-only.
Beta
Paper owners who already run voice chat, 15+ daily players. You use it and report what breaks.
- You are the owner
- Voice chat is in regular use
- Honest bug reports
Questions
Do you keep recordings?
No. Audio is processed in memory and discarded. Only the flagged sentence text stays, for 30 days. There is no player opt-out yet — disclosure is a join message.
When can I buy it?
Not yet. Listing news goes to Discord first. BuiltByBit is the launch storefront.