How Emotion Images and TTS Work in miku.gg

miku.gg bots can show emotion through images and speak through text-to-speech (TTS). According to the miku.gg documentation, emotion analysis is handled locally with SBERT embeddings, and the platform currently supports three TTS services: Azure TTS, NovelAI, and ElevenLabs. This applies to anyone running miku.gg from the source code or trying the live demo — no separate emotion or voice plugin is mentioned as required.

What the feature does

The documentation lists "Emotion images and TTS" as a core feature with two distinct parts:

• Emotion images — your bot displays an image that reflects its current emotional state.
• TTS — your bot has a voice, converting its text replies into spoken audio.

Both are described as built-in capabilities of the platform rather than optional add-ons.

How emotion detection works

Emotion analysis runs on local SBERT embeddings. In practice this means the emotion behind a bot's message is inferred from sentence-level embeddings computed on your own machine, rather than by sending text to an external emotion-classification API. The detected emotion then maps to the image the bot shows.

This matters if you care about latency or privacy: the emotion step stays local, while the TTS step depends on whichever external service you configure.

Supported TTS services

The documentation names three TTS providers:

Each is an external service, so using one means your bot's text is sent to that provider to generate audio. The docs do not state pricing, quotas, or authentication details for any of them — check each provider directly for those terms.

What you need to know before setting it up

• Emotion images and TTS are separate concerns. Emotion detection is local (SBERT); voice generation is external (one of the three TTS services). You can reason about and configure them independently.
• The docs describe the capability, not a step-by-step setup. The Overview page confirms the feature exists and lists the supported services, but does not walk through configuration. For exact setup steps, refer to the source code and the rest of the documentation site.
• miku.gg itself is open source and free to use, and you can either download the source and run your own chatbot or use the live demo. That statement covers the platform — it does not mean the third-party TTS providers are free.
• API compatibility is separate from TTS. miku.gg works with OpenAI-like APIs (the docs mention Aphrodite, TextGen WebUI, and Tabby). That governs which model backend you run, not which voice your bot uses.

Quick decision guide

• Want a bot that reacts visually to conversation tone without sending text to an emotion API? The local SBERT setup is the relevant part.
• Want your bot to speak? Pick one of Azure TTS, NovelAI, or ElevenLabs and configure it; the docs list these as the current options.
• Want to try it before running anything locally? The bot directory and live demo let you try bots without running the platform yourself.

For the authoritative and most current list of supported services and setup details, start from the miku.gg documentation Overview and follow the linked pages.