The easiest and most common way to obtain the ggml-medium.bin model is by using the download-ggml-model.sh script that comes with the whisper.cpp repository. From the command line, navigate to the models/ folder within your whisper.cpp directory and run the script:
It is important to note that the original GGML format is considered and has been superseded by its successor, GGUF (GGML Universal File). The primary software that popularized GGML, llama.cpp , officially dropped support for the GGML format on August 21st, 2023 .
Requires roughly 2 GB to 4 GB of available system memory or video memory. Parameters: ~769 Million. ggml-medium.bin
: Consider using the ggml-medium-q5_0.bin or q4_0 variants. These quantized versions compress the file size and speed up processing with almost zero loss in accuracy.
ggml-medium.bin is more than just a file; it is the enabler of high-accuracy, portable AI transcription. By bringing 769 million parameters into the efficient GGML environment, it allows users to unlock high-level speech-to-text technology on everyday consumer hardware. The easiest and most common way to obtain the ggml-medium
The most common way to utilize this file is through , the C++ port of Whisper.
The "medium" model is often considered the "sweet spot" for users who need higher accuracy than the "base" or "small" models but cannot afford the massive hardware requirements of the "large" models. Requires roughly 2 GB to 4 GB of
You can generate these quantized files yourself using the ./quantize tool included in the whisper.cpp repository. Use Cases for the Medium Model Why choose ggml-medium.bin over other sizes?