Text to Speech

Text to Speech lets Abolitus read model replies aloud after generation finishes.

This is useful when you want a more immersive reading flow, want to listen while multitasking, or want voice defaults that reinforce a character's identity.

What TTS Is and Is Not

TTS is a separate playback system.

It is not the same thing as text generation.

That means:

• you can like one route for writing and another for voice,
• speech billing comes from the speech provider rather than from Premium,
• and speech settings need their own configuration rather than assuming your text model route should automatically control everything.

Supported TTS Provider Types

Abolitus currently supports:

• ElevenLabs.
• OpenAI-compatible TTS routes.

TTS credentials are device-local.

Like model provider keys, they do not automatically sync across your devices.

What You Can Configure

The speech settings let you choose:

• the active speech provider,
• whether completed assistant replies auto-play,
• a default speech model,
• a default voice,
• and the endpoint for compatible routes.

Some character setups can also carry their own preferred voice and model values, which helps playback feel more consistent from one character to another.

Auto-Play Behavior

If auto-play is enabled, newly completed assistant replies can begin playback automatically.

This is good when:

• you want a hands-free reading rhythm,
• you are treating the conversation like an audio scene,
• or you want to stay in the flow without manually triggering playback every turn.

Keep it off when:

• you skim quickly,
• you regenerate often,
• or you do not want repeated audio interruptions during tuning.

ElevenLabs vs OpenAI-Compatible TTS

ElevenLabs

Choose ElevenLabs when:

• you already use ElevenLabs voices,
• voice identity matters a lot for immersion,
• or you want a polished speech-first route.

This setup is especially useful when you have a specific voice ID you already trust.

OpenAI-Compatible TTS

Choose an OpenAI-compatible route when:

• you want one speech stack that matches an existing compatible endpoint,
• you want more control over where the speech request is sent,
• or you want to point TTS at a local or self-managed compatible service.

This route is often the most flexible choice for users who already think in terms of compatible endpoints rather than branded voice libraries.

Voice and Model Defaults

The TTS system supports default model and default voice values per configured provider.

For compatible routes, a set of common preset voice names can be used quickly.

For ElevenLabs, custom voice IDs are especially relevant because many users already know exactly which voice they want.

The practical point is simple: you do not need to configure voice from scratch every single turn.

Privacy Boundary

TTS requests are made from the browser to your selected speech provider or endpoint.

That means the privacy boundary depends on where you point speech generation.

If you use a local compatible speech endpoint, the request path can stay under your own control.

If you use a cloud speech provider, that provider necessarily sees the text needed to synthesize audio.

This is the same basic rule as text generation: local is strongest for privacy, while cloud is strongest for convenience or service quality.

A Good Way to Start

1. Configure one provider only.
2. Pick one voice that is good enough.
3. Test with auto-play off first.
4. Turn auto-play on only if it improves your rhythm.
5. Add character-specific voice habits later.

This keeps the setup from becoming complicated before you know whether TTS is actually part of your daily workflow.

Best Practices

Match voices to roles, not to novelty

The best TTS setups usually use voices that reinforce readability and scene identity, not just voices that sound flashy in isolation.

Separate speech quality from writing quality

Do not assume your best writing route is automatically your best voice route.

Remember that TTS is billed separately

Speech quality, voice availability, and playback cost come from your selected TTS provider, not from Abolitus Premium.

Related Pages

• Read LLM Connections if you want the broader picture of local versus cloud provider boundaries.
• Read Billing and Premium to understand why Premium does not cover third-party voice usage.
• Read Wallpaper and Appearance if your goal is full presentation polish rather than audio alone.