The first slur read out loud on stream is the moment many streamers turn TTS off for good. It does not have to come to that. This guide explains how ControlPlay moderates every TTS message before it is spoken, which settings you control, and what to do live when someone still tries to ruin your stream.
Every message is checked before it is spoken
A ControlPlay TTS message passes two checks before any audio exists:
- Word filter. A local filter matches known slurs, profanity and your own blocked words, including the usual bypass tricks: letter substitutions, padded characters, spaced-out letters, split words and stretched spellings.
- Semantic check. An AI moderation layer then reads the message that would actually be spoken and scores it for hate, harassment, threats and similar categories. It catches what word lists cannot: context, innuendo and coded phrasing.
Only a message that passes both checks reaches the voice. A rejected message is never synthesized, so there is no race to mute it before viewers hear it.
One detail worth knowing: TTS messages are submitted through the Extension, not through your chat. Chat tools like AutoMod and Blocked Terms never see them, which is exactly why ControlPlay runs its own checks on the TTS text itself.
Mask or block: choose what happens on a hit
| Mode | What happens | Coins |
|---|---|---|
| Mask (default) | Detected words are replaced and the cleaned message still plays | Kept, since the message was delivered |
| Block | Any hit stops the message entirely | Returned to the viewer automatically |
Masking keeps the fun alive for borderline cases: the message plays, the offending word does not. When a hit cannot be cleanly localized to specific words, for example fully spaced-out letters or a long phrase, the message is blocked instead of played half-broken. In block mode, viewers do not lose Coins for a message that was never spoken, which keeps honest viewers experimenting.
What you can configure, and what stays on
- Everyday profanity. You decide whether casual swearing is spoken as-is. Many communities allow it; family-friendly channels keep it filtered.
- Your own blocked words. Add up to 100 entries, single words or phrases: your deadname, spoilers, recurring harassment terms, inside jokes gone stale. Every entry is automatically protected against the bypass tricks above, so do not bother adding leetspeak variants yourself.
- Always on. Slurs, high-risk categories and the semantic check cannot be disabled, by you or anyone else. That floor is deliberate.
Changes apply from the next message, so you can tighten the filter mid-stream after an incident without restarting anything.
ControlPlay speciality: your custom blocked words inherit the full bypass protection automatically. Add "spoiler" once, and s p o i l e r, sp0iler and spoiiiler are covered without any extra work.
Cost is moderation too
Free TTS invites drive-by spam waves. In ControlPlay, every TTS message is a Coin action with an amount and cooldown you set. That friction changes troll math: flooding your stream stops being free, while one well-timed message from a real viewer stays easy. If you are a small streamer wondering whether TTS is worth the risk, start with a higher Coin amount and a strict cooldown, then loosen both as your community proves itself.
The Twitch TTS setup guide covers voices, audio routing and the full configuration.
When something slips through live
No filter catches everything. New phonetic tricks and borderline phrasing can pass any automated system, so keep three live tools within reach:
- Timeout one viewer. From the OBS dock, pause a disruptive viewer for 15 or 60 minutes with one click, without disabling TTS for everyone. You can lift it early.
- The master switch. Turning TTS off entirely takes seconds and is the right call during a raid of trolls.
- Tighten and move on. Add the term to your blocked words, switch to block mode if needed, and let the filter handle repeats from the next message on.
After an incident, treat it like any chat moderation case: ban or timeout the account and check whether a clip or VOD segment needs deleting.
Set up moderated TTS
- Create a ControlPlay account, connect Twitch, install the Windows client and activate the Extension.
- Open the TTS module, pick a voice and choose mask or block mode.
- Decide on casual profanity and add your first custom blocked words.
- Set a Coin amount and cooldown that fit your channel size.
- Test with a harmless message and one deliberately filtered word from a viewer account.
Start stricter than you think you need. Loosening a filter is a pleasant announcement; tightening one after an incident is damage control.