jdvanaus-Fifth_Edition_Fantasy_Voices icon

Fifth Edition Fantasy Voices

The natural English voice for Fifth Edition Fantasy's Dungeon Master: Kokoro-82M run on your own PC (ONNX Runtime + MisakiSharp). Turn it on in game with: fef voice on

Last updated 3 days ago
Total downloads 35
Total rating 1 
Categories Mods Client-side Utility AI Generated Deep North Update
Dependency string jdvanaus-Fifth_Edition_Fantasy_Voices-1.0.0
Dependants 0 other packages depend on this package

This mod requires the following mods to function

jdvanaus-Fifth_Edition_Fantasy-1.7.0 icon
jdvanaus-Fifth_Edition_Fantasy

Fifth-edition tabletop rules for Valheim in 12 languages: 58 races, 13 classes, 176 spells, companions, realms, and a Dungeon Master who tells your story, runs scenes and speaks aloud.

Preferred version: 1.7.0

README

Fifth Edition Fantasy - Voices

The Dungeon Master of Fifth Edition Fantasy reads its narration aloud, and townsfolk speak their lines. This package holds the libraries for the natural English voice: Kokoro-82M, a small neural text-to-speech model that runs on your own PC's processor. Nothing you say or hear goes over the network.

How to use it

  1. Install this package with your mod manager (it brings Fifth Edition Fantasy with it).
  2. In game, open the console (F5) and type fef voice on - or set [Voice] Narration = true in the config.
  3. The first time, the mod downloads the voice model and its 54 voices from the model's own GitHub release (about 337 MB, once, checked by SHA-256) into BepInEx/config/FifthEditionFantasy/Voices. Windows' own voice reads aloud until it's ready.

Then every DM narration card, NPC line and townsperson's reply is spoken. The DM has its own voice (a British storyteller by default); every NPC gets a voice of their own, the same each time.

Command What it does
fef voice Status
fef voice on / off Turn narration on or off
fef voice dm bm_fable Change the DM's voice (fef voice list shows all 28 English voices)
fef voice speed 1.1 / volume 0.7 Speed and loudness
fef voice test Hello there Say something

Other languages don't need this package: they use Windows' own voices (Settings > Time & language > Speech adds more).

Speed: on a typical gaming PC a sentence is ready in well under the time it takes to say it (about 0.4x real time on a 6-core CPU). It uses half your CPU threads, up to six, only while it's speaking.

What's inside

File What Licence
Microsoft.ML.OnnxRuntime.dll, onnxruntime.dll, onnxruntime_providers_shared.dll, libonnxruntime.so Microsoft ONNX Runtime 1.20.1 (Windows and Linux) MIT
MisakiSharp.dll MisakiSharp 2.2.0 by Lyrcaxis - English text to phonemes (no eSpeak) Apache-2.0 (see licenses/)
compat/*.dll System.Buffers, System.Memory, System.Numerics.Vectors, System.Runtime.CompilerServices.Unsafe - only loaded if the game's runtime lacks them MIT

All files are the publishers' own, unchanged. Licence texts and notices are in licenses/. The Kokoro model (Apache-2.0, by hexgrad; ONNX export by thewh1teagle, MIT; training data credits: Koniwa CC BY 3.0, SIWIS CC BY 4.0) is downloaded separately on first use.