Text-to-Speech

Choose voices for NPCs, the narrator and the player, then test speech before changing in-game playback settings.

Set Up a Voice Connector

Start with PocketTTS, the initial local choice, or a service you already have working.

Install and start the local runtime in DwemerDistro, or configure hosted credentials in API Keys.
Open Configuration → TTS and create or edit a connector.
Select the service and fill in its address and voice/model options.
Save and assign it as the TTS Connector in your profile.
Preview a short line in TTS Studio, then test an NPC using that profile.

Player and Narration have their own settings. Changing an NPC profile does not automatically change their connectors.

Choose a Speech Service

Local speech

The recommended list includes PocketTTS, Chatterbox, XTTS FastAPI and OmniVoice. Each requires its corresponding runtime at the configured address.

Hosted speech

Inworld and Cartesia are also recommended options in the editor. Configure provider credentials and a voice available to your account. Check pricing and voice permissions before generating audio.

Other connectors

The editor also exposes Piper, xVASynth, MeloTTS, Mimic3, Azure, ElevenLabs, OpenAI, Kokoro, KoboldCPP, Zonos and Deepgram. Follow the selected driver's fields; not every service supports cloning or identical voice controls.

A provider-generated Voice ID is not interchangeable with another provider's ID. Preview using the connector that will actually speak in game.

Voice ID and samples

Open the NPC editor to inspect Voice ID. Biographies and race-based fallback choices can supply voice assignments, while cloning-capable services use voice samples where available.

Keep voice samples available on the server. Do not assume every provider accepts the same identifier or supports the same cloning features. Test the selected NPC voice after changing providers.

TTS Studio

Use TTS Studio to preview speech before testing it in game. Try a short sentence first. If the preview fails, check the connector, service, credentials and selected voice.

Voice filters change the sound of generated speech. Preview a filter before assigning it to a character; strong effects can make speech harder to understand.

Pronunciations

Open the Pronunciations tab in TTS Studio to add or edit pronunciation entries and use the play/test controls. Test Morrowind names in a full sentence with your chosen voice.

Only add replacements for words the provider gets wrong. Pronunciation varies by voice and provider, so a replacement that improves one setup can make another worse.

Vanilla dialogue and player speech

Enable vanilla dialogue speech in the in-game Scripts settings. With the default audio pause setting, it can play while the dialogue menu pauses the game. Selecting a new option or closing the dialogue menu stops the previous speech.

If player TTS is also enabled, the selected player dialogue option is spoken before the NPC’s response. Leave player TTS off if you only want to hear the NPC.

If speech is slow

Compare a short Studio preview with a short in-game reply. Long replies can play in sentence-sized pieces, but speech generation still depends on provider speed and available hardware.

If text arrives quickly but audio does not, check the speech service logs. If neither arrives, investigate the model request first.