Ripcy's Speech mod (TTS)

Ripcy's Speech Mod

A Comprehensive Voiceover Framework for Pillars of Eternity. This mod brings full Text-to-Speech (TTS) capabilities to every corner of Eora.

This work is heavily inspired by the original W40K: Rogue Trader Speech Mod by Osmodium. However, the core principle was changed to separate the Framework (which can be used with any Unity game via Unity Mod Manager) from the "Patcher Mod" (which provides the data and game-specific code).

Platform Support:

• []Windows: Fully supported.[/]

  []Linux / Proton: Supported, but currently English version only (Check FAQ how to install different piper models, Kokoro is multy-lang but supports limited number of languages, list of supported Kokoro languages).[/]

  []Steam Deck: I don't own a Steam Deck, so I haven't been able to test it there personally. However, I have tested it on native Linux and Proton and it works just fine, so it technically should work on the Deck without issues.[/]

  []OSX / Mac: Not supported. While it is technically possible to build this for Mac, I unfortunately just don't have the hardware or resources to build and test it.[/]

Heads up when updating from versions older than 3. 0

If you are updating from a version older than 3. 0, your mod settings will unfortunately be reset. I apologize for the inconvenience! I highly recommend taking a quick screenshot of your current settings in the mod manager before you update so you can easily put them back.

Which file should I download?

For Windows Users

I recommend using Windows_SAPI_Full version. It has best audio quality (means Natural Voices, check below) and speed, so you don't need to worry about the others. But you are free to try any of the optional releases.

For Linux and Proton (Steam Deck) Users

• []If you have Nexus Premium: Go ahead and download the appropriate Full version (Linux or Proton). It has everything you need right out of the box. (Aside from cloned voices)[/]

  []If you don't have Premium: Downloading huge files on the free Nexus tier takes forever. Instead, download the much smaller Lean version from the files tab, and then download the Data/data_v4. 0. 0.zip separately from Google Drive. Some optional releases like CUDA are only uploaded as Lean.[/]

  File naming explained:

  • []Full - Means plug-n-play version, simply extract it into mods folder and it should work (Aside from cloned voices they will always be at google drive because of licensing and what not)[/]

  • []Lean - Means this build does not contain Kokoro / Piper models and if you downloaded this version you have to download data.zip separately. Or download models manually from sherpa-onnx repo, explained in FAQ section.[/]

    []CPU - Means this build's TTS engines will run on your CPU.[/]

    []CUDA - Means this build's TTS engines will run on your GPU using Nvidia Cuda they are overall faster and less demanding on resources compared to CPU builds, but they come with Nvidia dll bloat (1. 5g of dlls...) Also they REQUIRE Nvidia GPU, you can check the list of all supported GPU's here. Might at some point try ZLUDA, but not sure.[/]

    []DIRECTML - This is Microsoft runtime similar to Nvidia Cuda it detects available hardware (CPU/GPU/NPU) and uses whichever is more suitable, it is generally faster than CPU and should support Amd and Intel GPU's. But because it is Microsoft technology they are stingy and ignore linux support. (Side note: Kokoro currently will still run on CPU brecause of the standing DirectML bug....)[/]

    []LIGHT - This build does NOT contain any TTS engines, it specifically made to play cloned audio, cloned audio is not included and should be downloaded separately from Google Drive.[/]

  ---

  Important Notes for Sherpa Engine / Linux / Proton Users

  If you are using the Sherpa engine (which includes Linux and Proton users relying on it), keep a few techni