CREPE (tiny) needs about 0.5 GB of memory at F16 (a 0.0 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 6 quants, from 0.0 GB (Q4_K) to 0.0 GB (F16).
CREPE (tiny) – a lightweight pitch estimator in GGUF format
CREPE (tiny) is a convolutional pitch (F0) estimator that processes raw 16 kHz mono audio. It outputs 360 sigmoid‑valued activations per 10 ms frame, spaced 20 cents apart. The model is available in several quantizations for low‑memory use.
Good at
- Fast inference on small GPUs
- Low memory footprint
- Direct raw waveform processing
Limits
- No language support
- No vision capability
- Requires 16 kHz mono audio input
Best for
- Real‑time pitch tracking
- Audio feature extraction
- Speech analysis tasks
Who it’s for. Developers needing a compact pitch estimator for real‑time or offline audio processing. The model is suitable for embedded systems or low‑resource environments.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K | 0.0 GB | 0.5 GB | Fits | Get file → | |
| Q8_0 | 0.0 GB | 0.5 GB | Fits | Get file → | |
| F16 | 0.0 GB | 0.5 GB | Fits | Get file → | |
| Q4_K | 0.0 GB | 0.5 GB | Fits | Get file → | |
| Q8_0 | 0.0 GB | 0.5 GB | Fits | Get file → | |
| F16 | 0.0 GB | 0.5 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_K (file size 0.01 GB) or Q8_0 (0.02 GB). For 4 GB VRAM, Q4_K (0.01 GB) is recommended. All variants estimate 0.5 GB memory at 8 k context, approximate.
What GPU do you need for CREPE (tiny)?
These figures add the file size, an fp16 KV cache at an 8K context window and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is F16 (0.0 GB file, about 0.5 GB in use).
Which CREPE (tiny) quant should you choose?
F16 (0.0 GB) is the middle option in this repository. The unquantized model. It gives original quality and needs the most memory, so it is mainly for conversion or reference.
If memory is tight, the smallest file is Q4_K at 0.0 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about CREPE (tiny)
How much VRAM does CREPE (tiny) need?
About 0.5 GB at F16 with an 8K context window. That is the 0.0 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run CREPE (tiny) on an 8 GB GPU?
Yes, with F16: the 0.0 GB file needs about 0.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of CREPE (tiny) GGUF?
It depends on the quant: Q4_K is 0.0 GB, Q8_0 is 0.0 GB, F16 is 0.0 GB, Q4_K is 0.0 GB, Q8_0 is 0.0 GB, F16 is 0.0 GB.
Where can I download CREPE (tiny) GGUF?
The files are on Hugging Face in the cstr/crepe-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does CREPE (tiny) use?
Hugging Face lists the license as mit. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run CREPE (tiny) on 4 GB VRAM?
Yes, the Q4_K quantization file is 0.01 GB and estimates 0.5 GB memory usage at 8 k context.
Does CREPE support stereo audio?
No, the model requires 16 kHz mono audio input.
What is the recommended quantization for best fidelity?
Prefer f16 or q8_0; q4_k does not meet a 0.999 cosine bar and may misclassify pitch bins.