Voice engine 0.4.0

• --parent-pid <pid> makes the engine exit when the game exits, including after a crash. The Unity mod 0.4.0 passes it.

Unity mod 0.4.0

• The game starts the voice engine from Mods/VoiceBox/server and stops it when you quit. No more starting it by hand.
• The HUD shows "starting the voices..." while the engine loads, then "voices ready".

Install: copy voicebox.dfmod into DaggerfallUnity_Data/StreamingAssets/Mods/, and the server folder to Mods/VoiceBox/server (setup). Needs voice engine 0.4.0.

Voice engine 0.3.1

• --models <folder> sets where downloaded models are kept.
• The log says when the first start is downloading the model (about a minute).
• Install notes give real disk use: about 5 GB for the GPU environment, plus 320 MB for the model.

Setup: server/README.md

Voice server 0.3.0

First public release of the voice server used by every VoiceBox edition.

• Kokoro (default) and Supertonic engines, each in its own virtual environment.
• Sentence-first streaming, a priority queue, a sentence cache, and cancel on disconnect.
• OpenAI-compatible POST /v1/audio/speech.

Setup: server/README.md

Unity mod 0.3.1

• On-screen messages and settings call it the voice engine, and say plainly when it isn't running.

Install: set up the voice engine, then copy voicebox.dfmod into DaggerfallUnity_Data/StreamingAssets/Mods/. Updating from 0.3.0: replace the file. Your voice tags and settings are kept.

Unity mod 0.3.0

First public release, for Daggerfall Unity 1.1.1.

Install

1. Set up and start the voice server.
2. Copy voicebox.dfmod into DaggerfallUnity_Data/StreamingAssets/Mods/.
3. Enable VoiceBox in the mod list and load a save.

In this release

• Dialogue cards, quest narration, world notices, other mods' popups, menu prompts, the talk window and HUD messages. Each can be set to Auto, Icon only or Off.
• Each mod's popups can be set separately in the VoiceBox window (Ctrl+F7, or the Esc menu).
• Voice tags: NPC, Narrator, You and Never, plus character tags (merchant, guard, noble…).
• Volume up to 300%, with a soft limiter.

This is a pre-release because the voice server is still a manual Python setup. A single download with the voice engine included is planned for 1.0. See the roadmap.