0.2.31

Higgs Audio v3 Studio v0.2.31

Windows and Linux release. This version adds full Linux CUDA support alongside easier engine loading, clearer dependency diagnostics, and smoother setup for users who do not have the full CUDA Toolkit installed.

Downloads

Windows

Higgs Audio v3 Studio_0.2.31_x64-setup.exe
Windows NSIS installer. Use this for normal installation.

Higgs Audio v3 Studio_0.2.31_x64_en-US.msi
Windows MSI installer. Use this if you prefer MSI-based installation or deployment.

Higgs Audio v3 Studio_0.2.31_x64_Portable.zip/
Portable non-installer folder. Keep the .exe and resources/ folder together.

Linux

Higgs Audio v3 Studio_0.2.31_amd64.deb
DEB package for Ubuntu/Debian. Install with sudo apt install ./Higgs-Audio-v3-Studio_0.2.31_amd64.deb.

Higgs Audio v3 Studio_0.2.31_amd64.AppImage
Portable AppImage for any Linux distro. Make executable with chmod +x and run.

Higgs Audio v3 Studio-0.2.31-1.x86_64.rpm
RPM package for Fedora/RHEL.

Requirements

Windows

• Windows x64
• NVIDIA GPU, recommended RTX 30xx / 40xx / 50xx
• CUDA 13 compatible NVIDIA driver
• Recommended VRAM:
  • q4_k_m: 8 GB VRAM
  • q5_k: 9 GB VRAM
  • q6_k: 10 GB VRAM
  • q8_0: 12 GB VRAM
  • bf16: 16 GB VRAM

Linux

• Ubuntu 24.04 LTS x64 (recommended) or other Debian-based distro
• NVIDIA GPU, recommended RTX 30xx / 40xx / 50xx
• CUDA 13 compatible NVIDIA driver
• Same VRAM requirements as Windows

What Is New In v0.2.31

Linux Support (New)

• Added native CUDA engine for Linux (libaudiocpp_engine.so) covering sm_86 (RTX 30xx), sm_89 (RTX 40xx), and sm_120a (RTX 50xx).
• Engine package bundles CUDA 13 runtime libraries (libcudart, libcublas, libcublasLt) with rpath=$ORIGIN, so users only need an NVIDIA driver installed — no CUDA Toolkit required.
• Fixed NVML telemetry: now loads libnvidia-ml.so.1 so GPU VRAM, power, and utilization appear in the hardware monitor on Linux.
• Disabled WebKitGTK DMABUF renderer to enable GPU-accelerated compositing on NVIDIA drivers.
• Replaced CSS zoom UI scaling with transform: scale() for smooth slider dragging on WebKitGTK.
• Fixed dark mode dropdown styling on WebKitGTK with color-scheme, appearance: none, and custom arrow SVGs.
• Added PNG icons for Linux desktop/DEB/RPM/AppImage bundling (generated from existing .ico).
• Added platform-aware engine download: Linux fetches from Hugging Face engines_linux/.
• Updated UI labels from "DLL" to "libraries" or "Files" on Linux.

Windows Maintenance

• Added clearer Windows engine dependency validation before loading audiocpp_engine.dll.
• Added detection for the actual current engine dependencies:
  • nvcuda.dll
  • cublas64_13.dll
  • cublasLt64_13.dll
  • MSVCP140.dll
  • VCOMP140.DLL
  • VCRUNTIME140.dll
  • VCRUNTIME140_1.dll
• Removed false required checks for CUDA DLLs not imported by this engine build.
• Updated the engine download button to Download Engine DLLs.
• Download Engine DLLs now downloads the engine package from the Hugging Face engines/ folder instead of only downloading audiocpp_engine.dll.
• The app can now use either system-installed CUDA/MSVC runtime DLLs or DLLs downloaded into the app engine folder.
• Improved engine loader search behavior so the loader uses the same dependency search folders checked by the diagnostic validator.
• Updated Hugging Face manifest/checksum layout for the full engine DLL package and Linux engine package.
• Updated README and README_ZH with the new engine package layout and Linux build/install instructions.
• Added multi-speaker preflight validation before generation starts.
• Added multi-speaker line progress text such as Line 1/42, Line 2/42, etc.
• Added clearer error handling when a multi-speaker line points to a missing speaker or a speaker without a reference voice.

What Is Included

• Tauri desktop UI (Windows and Linux)
• Ported Higgs Audio v3 C++/CUDA engine (Windows DLL, Linux .so)
• Higgs config/tokenizer asset files
• Model download UI
• Engine package download UI (platform-aware)
• Whisper model selection/download UI
• Voice clone workflow
• Continue speech workflow
• Multi-speaker workflow
• Speaker Gallery
• WAV/MP3 save support
• Local API server
• API streaming support
• API examples for curl, Python, JavaScript, and PowerShell
• API Command Centre log viewer
• Engine dependency repair diagnostics
• NVIDIA hardware telemetry (VRAM, GPU load, power, RAM, history graph)
• Bundled external-browser links for GitHub, releases, Hugging Face, NVIDIA driver/CUDA, VC++ runtime, and Whisper model selection

What Is Not Included

The large Higgs GGUF model weights are not bundled with the installer or portable build. Download them from the app or from the Hugging Face model repository.

The NVIDIA driver is not bundled. Users still need a CUDA 13 compatible NVIDIA driver installed.

nvcuda.dll / libcuda.so is not bundled and should not be manually copied. It is provided by the NVIDIA driver.

Model And Runtime Files

Model and engine runtime files are hosted here:

https://huggingface.co/drbaph/Higgs-Audio-v3-Studio

Expected Hugging Face layout:

/
├─ manifest.json
├─ checksums/
│  └─ SHA256SUMS.txt
├─ engines/                          # Windows engine package
│  ├─ audiocpp_engine.dll
│  ├─ cublas64_13.dll
│  ├─ cublasLt64_13.dll
│  ├─ MSVCP140.dll
│  ├─ VCOMP140.DLL
│  ├─ VCRUNTIME140.dll
│  └─ VCRUNTIME140_1.dll
├─ engines_linux/                    # Linux engine package
│  ├─ libaudiocpp_engine.so
│  ├─ libcudart.so.13
│  ├─ libcublas.so.13
│  └─ libcublasLt.so.13
└─ models/
   ├─ higgs-q4_k_m/
   │  └─ q4_k_m.gguf
   ├─ higgs-q5_k/
   │  └─ q5_k.gguf
   ├─ higgs-q6_k/
   │  └─ q6_k.gguf
   ├─ higgs-q8_0/
   │  └─ q8_0.gguf
   └─ higgs-bf16/
      └─ bf16.gguf

Available Higgs Models

• q4_k_m: smaller model, recommended for 8 GB VRAM
• q5_k: balanced K-quant model, recommended for 9 GB VRAM
• q6_k: higher-quality K-quant model, recommended for 10 GB VRAM
• q8_0: recommended default, best quality/speed balance, recommended for 12 GB VRAM
• bf16: highest fidelity testing model, recommended for 16 GB VRAM

Windows Setup

Installer

1. Download Higgs Audio v3 Studio_0.2.31_x64-setup.exe.
2. Run the installer.
3. Open Higgs Audio v3 Studio.
4. If the engine/runtime DLLs are missing, click Download Engine DLLs.
5. Download/select a Higgs model if needed.
6. Load the engine, then load the model.
7. Pick a workflow and generate audio.

Portable

Download or copy the full Higgs Audio v3 Studio Portable folder.

Keep this structure intact:

Higgs Audio v3 Studio Portable/
├─ Higgs Audio v3 Studio.exe
└─ resources/
   ├─ engine/
   │  └─ audiocpp_engine.dll
   └─ higgs-assets/
      └─ higgs-audio-v3-tts-4b/
         ├─ config.json
         ├─ tokenizer.json
         ├─ tokenizer_config.json
         ├─ higgs_audio_v2_tokenizer_config.json
         └─ chat_template.jinja

If the portable build reports missing CUDA/MSVC runtime DLLs, open the app and click Download Engine DLLs. The downloaded runtime package is stored in the user-writable app engine folder.

Linux Setup

DEB (Ubuntu/Debian)

sudo apt install ./Higgs-Audio-v3-Studio_0.2.31_amd64.deb

Launch from the app menu or run higgs-audio-studio from terminal.

AppImage (portable)

chmod +x Higgs-Audio-v3-Studio_0.2.31_amd64.AppImage
./Higgs-Audio-v3-Studio_0.2.31_amd64.AppImage

RPM (Fedora/RHEL)

sudo dnf install ./Higgs-Audio-v3-Studio-0.2.31-1.x86_64.rpm

After launching

1. Click Download Engine Files to fetch the Linux CUDA engine package.
2. Download or browse to a Higgs model folder.
3. Load the engine, then load the model.
4. Pick a workflow and generate audio.

The engine .so files download to ~/.higgs-audio-v3-studio/engine/ and include all required CUDA runtime libraries. Users only need an NVIDIA driver — no CUDA Toolkit install required.

Notes

• The app can use runtime libraries already installed on the system.
• The app can also use runtime libraries downloaded from the Hugging Face engine folders.
• Users still need an NVIDIA driver. The app cannot bundle or replace nvcuda.dll / libcuda.so.
• External links open in the system browser.
• If the API Command Centre is popped out, it can remain visible while the main app is minimized to tray.

0.2.3

Higgs Audio v3 Studio v0.2.3

Windows desktop maintenance release of Higgs Audio v3 Studio, a Rust/Tauri app for local Higgs Audio v3 TTS inference through a ported native C++/CUDA engine.

Downloads

Higgs Audio v3 Studio_0.2.3_x64-setup.exe

Windows NSIS installer. Use this for normal installation.

Higgs Audio v3 Studio_0.2.3_x64_en-US.msi

Windows MSI installer. Use this if you prefer MSI-based installation or deployment.

Higgs Audio v3 Studio_0.2.3_x64 Portable.zip/

Portable non-installer folder. Keep the .exe and resources/ folder together.

Requirements

• Windows x64
• NVIDIA GPU: 30xx, 40xx, or 50xx
• CUDA 13 runtime/driver support

Recommended VRAM

Model	Recommended VRAM
q4_k_m	8 GB
q5_k	9 GB
q6_k	10 GB
q8_0	12 GB
bf16	16 GB

What Is New In v0.2.3

• Enabled F16 decode KV cache by default in the Higgs CUDA engine.
• Reduced decode KV cache VRAM pressure compared with the previous F32 cache path.
• Fixed the F16 KV cache crash caused by feeding F16 rows into ggml_set_rows.
• Kept a safe F32 fallback path for diagnostics.
• Disabled unsupported mixed K/V cache modes so they no longer hit CUDA flash-attention fatal errors.
• Rebuilt the CUDA engine DLL for NVIDIA 30xx, 40xx, and 50xx targets.
• Refreshed the bundled engine DLL in installer, MSI, and portable packages.
• Updated Hugging Face manifest and SHA256SUMS.txt for the corrected engine DLL.
• Kept the app version at v0.2.3.

What Is Included

• Tauri desktop UI
• Ported Higgs Audio v3 C++/CUDA engine DLL
• Higgs config/tokenizer asset files
• Model download UI
• Whisper model selection/download UI
• Voice clone workflow
• Continue speech workflow
• Multi-speaker workflow
• Speaker Gallery
• WAV/MP3 save support
• Local API server
• API streaming support
• API examples for curl, Python, JavaScript, and PowerShell
• API Command Centre log viewer
• Bundled external-browser links for GitHub, releases, Hugging Face, and Whisper model selection

What Is Not Included

The large Higgs GGUF model weights are not bundled with the installer or portable build. Download them from the app or from the Hugging Face model repository.

Model files are hosted here:

https://huggingface.co/drbaph/Higgs-Audio-v3-Studio

Available Higgs Models

Model	Description
q4_k_m	Smaller model, recommended for 8 GB VRAM
q5_k	Balanced K-quant model, recommended for 9 GB VRAM
q6_k	Higher-quality K-quant model, recommended for 10 GB VRAM
q8_0	Recommended default, best quality/speed balance, recommended for 12 GB VRAM
bf16	Highest fidelity testing model, recommended for 16 GB VRAM

Installer Setup

1. Download Higgs Audio v3 Studio_0.2.3_x64-setup.exe.
2. Run the installer.
3. Open Higgs Audio v3 Studio.
4. Download/select the engine and model if needed.
5. Load the engine, then load the model.
6. Pick a workflow and generate audio.

Portable Setup

Download or copy the full Higgs Audio v3 Studio Portable/ folder.

Keep this structure intact:

Higgs Audio v3 Studio Portable/
├─ Higgs Audio v3 Studio.exe
└─ resources/
   ├─ engine/
   │  └─ audiocpp_engine.dll
   └─ higgs-assets/
      └─ higgs-audio-v3-tts-4b/
         ├─ config.json
         ├─ tokenizer.json
         ├─ tokenizer_config.json
         ├─ higgs_audio_v2_tokenizer_config.json
         └─ chat_template.jinja

Hugging Face Update Files

Upload these files to the Hugging Face model repository:

engines/audiocpp_engine.dll
manifest.json
checksums/SHA256SUMS.txt

Notes

• The bundled engine DLL in this release is the corrected CUDA engine with F16 decode KV cache enabled by default.
• The app downloads model weights separately into a user-writable model folder.
• External links open in the system browser.
• If the API Command Centre is popped out, it can remain visible while the main app is minimized to tray.

0.2.0

Higgs Audio v3 Studio v0.2.0

Second Windows desktop release of Higgs Audio v3 Studio, a Rust/Tauri app for local Higgs Audio v3 TTS inference through a ported native C++/CUDA engine.

Downloads

Higgs Audio v3 Studio_0.2.0_x64-setup.exe

Windows NSIS installer. Use this for normal installation.

Higgs Audio v3 Studio_0.2.0_x64_en-US.msi

Windows MSI installer. Use this if you prefer MSI-based installation or deployment.

Higgs Audio v3 Studio _0.2.0_x64 Portable/

Portable non-installer folder. Keep the .exe and resources/ folder together.

Requirements

• Windows x64
• NVIDIA GPU (30xx, 40xx, 50xx)
• CUDA 13 runtime/driver support
• Recommended VRAM:
  • q4_k_m: 8 GB VRAM
  • q5_k: 9 GB VRAM
  • q6_k: 10 GB VRAM
  • q8_0: 12 GB VRAM
  • bf16: 16 GB VRAM

What Is New In v0.2.0

• Added live streaming playback during generation.
• Added API streaming endpoint with NDJSON progress/audio events.
• Added local API Command Centre with popout window and filters.
• Fixed API Command Centre popout ACL permission.
• Fixed API Info examples panel height/layout.
• Added unified generation queue handling for UI/API jobs.
• Added Speaker Gallery for reusable speaker identities.
• Added speaker ZIP import/export with selected-speaker export.
• Added saved speaker reference cache support using .hspkcache.
• Added Q5_K and Q6_K Higgs model options.
• Updated model manifest and checksums for Q5_K/Q6_K.
• Added model dropdown support for Q5_K and Q6_K.
• Improved portable package with updated bundled engine DLL and assets.
• Updated README and README_ZH for v0.2.0.

What Is Included

• Tauri desktop UI
• Ported Higgs Audio v3 C++/CUDA engine DLL
• Higgs config/tokenizer asset files
• Model download UI
• Whisper model selection/download UI
• Voice clone workflow
• Continue speech workflow
• Multi-speaker workflow
• Speaker Gallery
• WAV/MP3 save support
• Local API server
• API streaming support
• API examples for curl, Python, JavaScript, and PowerShell
• API Command Centre log viewer
• Bundled external-browser links for GitHub, releases, Hugging Face, and Whisper model selection

What Is Not Included

The large Higgs GGUF model weights are not bundled with the installer or portable build. Download them from the app or from the Hugging Face model repository.

Model files are hosted here:

https://huggingface.co/drbaph/Higgs-Audio-v3-Studio

Available Higgs Models

• q4_k_m: smaller model, recommended for 8 GB VRAM
• q5_k: balanced K-quant model, recommended for 9 GB VRAM
• q6_k: higher-quality K-quant model, recommended for 10 GB VRAM
• q8_0: recommended default, best quality/speed balance, recommended for 12 GB VRAM
• bf16: highest fidelity testing model, recommended for 16 GB VRAM

Installer Setup

1. Download Higgs Audio v3 Studio_0.2.0_x64-setup.exe.
2. Run the installer.
3. Open Higgs Audio v3 Studio.
4. Download/select the engine and model if needed.
5. Load the engine, then load the model.
6. Pick a workflow and generate audio.

Portable Setup

Download or copy the full Higgs Audio v3 Studio Portable folder.

Keep this structure intact:

Higgs Audio v3 Studio Portable/
├─ Higgs Audio v3 Studio.exe
└─ resources/
   ├─ engine/
   │  └─ audiocpp_engine.dll
   └─ higgs-assets/
      └─ higgs-audio-v3-tts-4b/
         ├─ config.json
         ├─ tokenizer.json
         ├─ tokenizer_config.json
         ├─ higgs_audio_v2_tokenizer_config.json
         └─ chat_template.jinja

Notes

• The bundled engine DLL in this release is the updated CUDA engine.
• The app downloads model weights separately into a user-writable model folder.
• External links open in the system browser.
• If the API Command Centre is popped out, it can remain visible while the main app is minimized to tray.

0.1.0

Higgs Audio v3 Studio v0.1.0

Initial Windows desktop release of Higgs Audio v3 Studio, a Rust/Tauri app for local Higgs Audio v3 TTS inference through a ported native C++/CUDA engine.

Downloads

• Higgs Audio v3 Studio_0.1.0_x64-setup.exe
  Windows NSIS installer. Use this for normal installation.

• Higgs Audio v3 Studio_0.1.0_Portable/
  Portable non-installer folder. Keep the .exe and resources/ folder together.

• higgs-assets-resources-v0.1.0.zip
  Standalone Higgs asset bundle containing resources/higgs-assets/.... This is only needed if you are manually repairing or preparing a portable folder.

Requirements

• Windows x64
• NVIDIA GPU
• CUDA 13 runtime/driver support
• Recommended VRAM:
  • q4_k: 8 GB VRAM
  • q8: 12 GB VRAM
  • bf16: 20 GB VRAM

What Is Included

• Tauri desktop UI
• Ported Higgs Audio v3 C++/CUDA engine DLL
• Higgs config/tokenizer asset files
• Model download UI
• Whisper model selection/download UI
• Voice clone, continue speech, and multi-speaker workflow UI
• WAV/MP3 save support
• Bundled external-browser links for GitHub, releases, Hugging Face, and Whisper model selection

What Is Not Included

The large Higgs GGUF model weights are not bundled with the installer or portable build. Download them from the app or from the Hugging Face model repository.

Installer Setup

1. Download Higgs Audio v3 Studio_0.1.0_x64-setup.exe.
2. Run the installer.
3. Open Higgs Audio v3 Studio.
4. Download/select the engine and model if needed.
5. Load the engine, then load the model.

Portable Setup

1. Download or copy the full Higgs Audio v3 Studio Portable folder.

2. Keep this structure intact:

   Higgs Audio v3 Studio Portable/
   ├─ Higgs Audio v3 Studio.exe
   └─ resources/
      ├─ engine/
      │  └─ audiocpp_engine.dll
      └─ higgs-assets/
         └─ higgs-audio-v3-tts-4b/
            ├─ config.json
            ├─ tokenizer.json
            ├─ tokenizer_config.json
            ├─ higgs_audio_v2_tokenizer_config.json
            └─ chat_template.jinja