Downloads · 30 days
0
cstr/whisper-large-v3-turbo-german-GGUF
whisper-large-v3-turbo-german-GGUF is a automatic speech recognition model from cstr. Use it when you need speech turned into text. It is set up for transformers. The card lists the license as apache-2.0.
GGML conversions and quantisations of primeline/whisper-large-v3-turbo-german for use with CrispStrobe/CrispASR or any whisper.cpp-compatible tool.
Downloads · 30 days
0
Access
Public
Updated Apr 18, 2026
Repo size
2.7 GB
Likes
0
Public
Click a slice to open those files.
.bin2.7 GB · 100%
From the Hugging Face model README
GGML conversions and quantisations of primeline/whisper-large-v3-turbo-german for use with CrispStrobe/CrispASR or any whisper.cpp-compatible tool.
| File | Quant | Size | Notes |
|---|---|---|---|
ggml-model.bin | F16 | 1.6 GB | Original conversion, full precision |
ggml-model-q5_0.bin | Q5_0 | 548 MB | Good quality/size tradeoff |
ggml-model-q4_k.bin | Q4_K | 453 MB | Smallest, fastest on CPU |
All variants produce correct German transcription on test audio. Q4_K is recommended for CPU deployment.
primeline/whisper-large-v3-turbo-germanThe "turbo" variant uses only 4 decoder layers (vs 32 in large-v3), making it ~3x faster at inference with minimal quality loss for German.
# Build CrispASR
git clone https://github.com/CrispStrobe/CrispASR && cd CrispASR
cmake -S . -B build && cmake --build build -j8
# Transcribe German audio
./build/bin/crispasr -m ggml-model-q4_k.bin -f german_audio.wav -l de
# With subtitles
./build/bin/crispasr -m ggml-model-q4_k.bin -f german_audio.wav -l de -osrt --split-on-punct
Converted from the original HuggingFace model using whisper.cpp's convert-h5-to-ggml.py, then quantised with whisper-quantize:
python models/convert-h5-to-ggml.py primeline/whisper-large-v3-turbo-german . models
whisper-quantize ggml-model.bin ggml-model-q5_0.bin q5_0
whisper-quantize ggml-model.bin ggml-model-q4_k.bin q4_k