01 · Product overview Livestreaming · local voice interaction · Windows

Say the cue.
Change the scene.

LivePet lets a host say a fixed voice command to switch a product card, trigger a character action, play a sound, or show a caption. Recognition, assets, and show settings stay on the local computer.

480msdefault audio chunk
9 · 27built-in actions · voice cues
4characters on screen
0audio uploaded to cloud

Common usesRetailBeautyFoodGamingEducationEntertainment

Scroll for the next section

02 · At a glance

A cue becomes a confirmed visual change

Microphone input, local recognition, command confirmation, and rendering happen on the same Windows machine.

A host says a fixed voice cue while a vertical livestream product card changes — Voice cue → confirmation → product card

Say the cueUse the default Windows microphone. Add synonyms and common recognition variants for each rule.
Confirm the sentenceStreaming recognition provides feedback, while context confirmation decides whether to trigger.
Render and acknowledgeThe target asset is decoded, checked, drawn, and acknowledged before the state advances.

A product card with a green confirmation mark and rejected trigger bubbles fading away — Only a confirmed command reaches the scene.

Images are product illustrations. The installed client remains the source of truth for the final interface.

03 · How it works

Three steps, one local loop

LivePet does not switch the scene just because it heard a keyword. The full command must be confirmed first.

1

Configure a rule

Bind a phrase such as “show card three” to a card or character action. Add alternate phrases when needed.

2

Recognize locally

Streaming output gives the host feedback; contextual confirmation decides whether the action is valid.

3

Commit the visual

The asset is loaded, validated, rendered, and acknowledged before it replaces the last confirmed scene.

Example: 30 seconds in a product livestream

Show card threeSwitch to the third product image
Show detailOpen the material close-up
Next cardAdvance through the card list
Model turnPlay a character or motion card

The phrases above are examples. Each show defines its own rules and cooldowns.

A microphone wave becoming three floating product cards — One voice cue can map to one visual action.

04 · Scenarios

Built for live moments where hands are busy

Each scenario can use its own cards, phrases, characters, and show settings.

Retail

Hands stay on product

Move from hero image to size, material, and detail views while presenting.

Show card threeProduct hero
Show detailClose-up
Next cardList loop

Beauty

Compare quickly

Switch between swatches, ingredients, and before-and-after visuals without breaking rhythm.

Show swatchColor view
Show ingredientsFormula card
Back to heroMain product

Education

Follow the lesson

Advance slides, diagrams, and examples as the presenter moves through a lesson.

Show outlineCourse outline
Show exampleCase slide
Next cardLesson flow

One show, one repeatable setupKeep product cards, voice rules, sound effects, captions, and rehearsal data together in a local program. Changes apply to the next run without rebuilding the whole broadcast.

05 · Capabilities

A practical control layer for the show

LivePet covers the repeated operational work around a voice-triggered visual interaction.

Voice rules

Fixed phrases, synonyms, and recognition variants
Per-rule cooldown and action sequences
Keyboard and simulated event testing

Visual output

16:9, 9:16, 1:1, and custom canvas sizes
Cards, characters, sounds, and captions
Direct output window for livestream software

Operations

Preflight checks and rehearsal mode
Backups, diagnostics, and support exports
Local analytics without audience identity data

Platform events such as gifts, comments, and follows are local test events until a formal platform integration is enabled.

06 · Privacy and setup

Local by default. Ready in three steps.

The core recognition and rendering loop is designed to run on the host computer.

Recognition stays local

The model and inference run on the machine. Microphone audio and voice cues do not need to be uploaded.

Show data stays local

Programs, settings, assets, rehearsal history, and analytics are stored in the local data directory.

Diagnostics are redacted

Support exports omit activation codes and keep device identifiers partially masked.

Analytics avoid content

Local statistics contain dates, outcomes, sources, and rule IDs, not recordings or audience identity.

Credentials are protected

OBS credentials and order keys use Windows DPAPI instead of plaintext configuration values.

Start with a preflight

Install the client, connect the output window to your livestream software, configure cues, and rehearse before going live.

Install and openRun the Windows client and use the first-run wizard to check the microphone.
Connect the canvasOpen the output window and add it as a window source in your livestream software.
Configure and rehearseImport cards, bind cues, run preflight, and start the show once the checks pass.