Companion

Pixely is a separate AI co-host layer: a 2D avatar, spoken reactions and live-chat interactions designed to stay lightweight while you stream.

Before you start

Enable Companion and add its separate Browser Source from Overlays. Configure an available AI provider and voice before expecting spoken replies.

Bring your Companion on stream

In Overview, choose the name, personality, language and avatar. In Voice, install/select the local voice or configure your chosen external provider.
Enable one reaction, such as responding to mentions. Test that interaction before adding greetings, automatic reactions or Auto Talk.
To talk through your computer microphone, open Interaction, enable it, install the speech-recognition package and set a Talk hotkey. Press once to start and again to send; use the preview to test.

Check the result

You see and hear the expected response. If the avatar is missing, check the Companion source and visibility mode. If there is no answer, check AI/voice setup and any limit or error. If the microphone is silent, check the selected PC input and permissions.

Settings and further help

OVERVIEW

Name, personality & avatar

Choose the Companion name, personality, language, display mode and included avatar — or supply a matching custom avatar set.

REACTIONS

Mentions + automatic reactions

Respond when viewers address the Companion by name, greet selected viewers, react to stream events and use Auto Talk after quiet periods.

VOICE

Local or external voice

Install the free local voice package for supported PixelyChat languages, or connect a supported external voice provider such as ElevenLabs when configured.

ADVANCED

Visibility & stream behavior

Use Always visible, Only when active or Audio-only, with independent horizontal and vertical stream placement.

Companion reactions — CompanionREACTIONS

Companion voice settings — CompanionVOICE

Companion advanced settings — CompanionADVANCED

Talk to your Companion.

Switch it on under Companion → Interaction. Press your Talk hotkey, say something and press it again. Your voice is turned into text on your PC, the Companion's AI platform writes the answer, and the Companion says it on stream a few seconds later. While you talk, the avatar shows a listening pose and stops what it was saying; waiting reactions come after your reply. The Companion Browser Source has to be in OBS.

Talk hotkeyPress once to start talking and again to send. It sends on its own after 30 seconds.

Cancel hotkeyOptional. Throws the recording away. It only takes the key while you're talking.

Speech RecognitionStandard (about 160 MB) for English and the major European languages, or Best accuracy (about 375 MB) for Japanese, Chinese, Korean, Arabic and other languages. Runs on your PC; the recording is never saved.

On-Screen IndicatorShows Listening and your hotkeys on your screen while you talk. Only you see it. Pick its position or turn it off.

PreviewTalk to your Companion right in the tab and see and hear the answer in both previews. Nothing is posted to chat.

Companion Interaction settings — CompanionINTERACTION

Chat and AI limit

Replies are also posted to Twitch and Kick chat while you're live when Send event reactions to chat is on in Advanced. With PixelyChat AI, replies have the same hourly limit as AI Chat.

When Bot AI and Companion overlap

Companion Mentions use the same AI response path rather than creating a duplicate response. Choose Companion Only or Companion + ChatBot for how that one direct reply is presented.