Whisper.cpp Installation Guide

Downloading whisper.cpp

Download Pre-built Executable (Recommended)

1. Go to the whisper.cpp releases page: https://github.com/ggerganov/whisper.cpp/releases

2. Download the latest version for Windows (look for files named like whisper-bin-x64.zip or whisper-blas-bin-x64.zip or similar)

3. Extract the ZIP file to your desired location (e.g., C:\Program Files\whisper.cpp\)

Downloading Models

In order to use whisper.cpp you need to download models to be used by whisper.

Where to Get Models

1. Go to the Hugging Face model repository: https://huggingface.co/ggerganov/whisper.cpp

2. Look for models with .bin extension in the "Files and versions" tab

Available Models (Recommended)

• tiny: ggml-tiny.bin (75 MB) - Fastest, lowest accuracy

• base: ggml-base.bin (142 MB) - Good balance

• small: ggml-small.bin (466 MB) - Better accuracy

• medium: ggml-medium.bin (1.5 GB) - High accuracy

• large: ggml-large-v3.bin (2.9 GB) - Highest accuracy

Download Steps

1. Choose the model that fits your needs (start with base or small for testing)

2. Click the download button next to the .bin file

3. Save the model file to your models folder

File Structure Setup

Recommended Folder Structure

whisper.cpp/
├── whisper-cli.exe          (executable)
├── models/              (models folder)
   ├── ggml-tiny.bin
   ├── ggml-base.bin
   ├── ggml-small.bin
   └── ggml-medium.bin

Installation Locations

• Executable path: C:\Program Files\whisper.cpp\whisper-cli.exe

• Models folder: C:\Program Files\whisper.cpp\models\

Application Configuration

In SubtitleNEXT, go to Options/Preferences, Speech to Text page, add Whisper (whisper.cpp) and configure these paths (assuming the above file paths are used, else change with your actual paths):

Executable

Set this to the location of whisper-cli.exe:

C:\Program Files\whisper.cpp\whisper-cli.exe

Model Path

Set this to the folder containing your downloaded models:

C:\Program Files\whisper.cpp\models\

Model

Select or write one of the downloaded models which you want to be used for this preset. For example:

ggml-large-v3.bin