kokoro voices needs about 0.5 GB of memory at its main file (a 0.0 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 7 quants, from 0.0 GB to 0.0 GB.
Kokoro voices – per‑speaker voice packs for the Kokoro‑82M family
Kokoro voices are small voice‑pack files (~510 KB) in GGUF format. They are designed for use with CrispStrobe/CrispASR alongside the Kokoro‑82M or German backbone models. Each pack contains a single F32 tensor voice.pack[max_phon, 1, 256].
Good at
- Small file size, no quantisation needed
- Direct passthrough loading
- Compatible with CrispASR and CrispStrobe
Limits
- No language packs for non‑native languages, falls back to ff_siwis
- Requires CrispASR/CrispStrobe for usage
- No vision support
Best for
- Adding voice packs to Kokoro‑82M setups
- German voice cascade in CrispASR
- Low‑memory voice loading
Who it’s for. Users of the Kokoro‑82M family who need per‑speaker voice packs for CrispStrobe or CrispASR. Ideal for developers integrating German or other language voice packs into existing Kokoro pipelines.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.0 GB | 0.5 GB | Fits | Get file → | |
| — | 0.0 GB | 0.5 GB | Fits | Get file → | |
| — | 0.0 GB | 0.5 GB | Fits | Get file → | |
| — | 0.0 GB | 0.5 GB | Fits | Get file → | |
| — | 0.0 GB | 0.5 GB | Fits | Get file → | |
| — | 0.0 GB | 0.5 GB | Fits | Get file → | |
| — | 0.0 GB | 0.5 GB | Fits | Get file → |
All voice packs are ~510 KB and use a single F32 tensor, so no quantisation is required. Loading is direct passthrough; memory usage is minimal and independent of quant level.
What GPU do you need for kokoro voices?
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 the main file (0.0 GB file, about 0.5 GB in use).
Which kokoro voices quant should you choose?
If memory is tight, the smallest file is the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about kokoro voices
How much VRAM does kokoro voices need?
About 0.5 GB at its main file 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 kokoro voices on an 8 GB GPU?
Yes, with the smaller files: 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 kokoro voices GGUF?
It depends on the quant: file is 0.0 GB, file is 0.0 GB, file is 0.0 GB, file is 0.0 GB, file is 0.0 GB, file is 0.0 GB and 1 more.
Where can I download kokoro voices GGUF?
The files are on Hugging Face in the cstr/kokoro-voices-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 kokoro voices use?
Hugging Face lists the license as apache-2.0. 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 Kokoro voices on a machine with 8 GB VRAM?
Yes, each voice pack is only ~510 KB, so VRAM usage is negligible. The model itself requires the Kokoro‑82M or German backbone for full operation.
Do I need to quantise the voice packs?
No. The voice packs are already in F32 format and are designed to load directly without quantisation.
What happens if I use a language without a native voice pack?
The system falls back to the ff_siwis pack, which is a French baseline non‑silence fallback.