Dialectic Text-to-Speech Services
Text-to-speech gives NPCs and the player audible voices. Dialectic can map Fallout voice types to installed samples for supported voice-cloning connectors.
Recommended
Use Configuration - TTS to select a connector, then open TTS Studio to test voice mappings and uploaded samples.
PocketTTS
PocketTTS is a fast local voice-cloning option and a good starting point for lower-latency gameplay. Install its DwemerDistro component and accept the model terms at Hugging Face.
Dialectic sends the resolved Fallout voice sample as the speaker reference when one is available.
Chatterbox
Chatterbox provides higher-quality local voice cloning with greater hardware requirements. Accept the model terms at Hugging Face before installing the component.
XTTS
XTTS supports multilingual voice cloning through the XTTS FastAPI service. Use TTS Studio to upload or refresh a sample and verify that the connector is synced before entering the game.
OmniVoice
OmniVoice is a locally hosted voice service managed through DwemerDistro. Its own configuration tool controls installed languages and generated voices.
Inworld
Inworld provides hosted character voices. Configure the required credentials and voice ID in the TTS connector.
Cartesia
Cartesia provides hosted low-latency speech. Connector settings control the model, voice, endpoint, and API key.
ElevenLabs
ElevenLabs uses hosted voices identified by voice ID. It does not use Dialectic's local Fallout sample cloning path in the same way as PocketTTS, Chatterbox, or XTTS.
Piper Text-to-Speech
Piper is a lightweight local synthesizer using installed fixed voice models. It is useful when voice cloning is not required.
Kokoro
Kokoro provides local speech from its supported voice set. Select a compatible voice in the connector configuration.
Fallout Voice Samples
Dialectic does not distribute voice audio extracted from Fallout. Its voice map points to samples already present in the installed game's Data/Sound/Voice tree and imports them when actors are encountered.
The Tools MCM page can request a broader voice-sample scan when an NPC still falls back to the connector's default voice.
Testing A Voice
Use the play control in TTS Studio. The standard test line is:
Patrolling the Mojave almost makes you wish for a Nuclear Winter.
A successful preview confirms server generation. In-game playback additionally depends on the plugin audio queue, cache download, actor mapping, lip animation, and subtitle settings.