fastpitch en needs about 0.8 GB of memory at Q4_K (a 0.2 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 0.1 GB (F16) to 0.2 GB (Q8_0).
FastPitch English TTS model with HiFi-GAN vocoder
FastPitch is a fully‑parallel transformer TTS model that generates mel spectrograms in a single forward pass. The model is combined with a HiFi‑GAN vocoder to produce 22050 Hz PCM audio. It is trained on LJSpeech and supports a single female English voice.
Good at
- Fast, non‑autoregressive synthesis
- High‑quality 22050 Hz audio
- Deterministic output
Limits
- Single speaker only
- Requires 22050 Hz vocoder
- No G2P phoneme conversion
Best for
- Real‑time TTS demos
- Batch text‑to‑speech conversion
- Embedded TTS applications
Who it’s for. Developers building text‑to‑speech systems that need fast, deterministic synthesis in English. The model is suitable for edge devices or servers with limited GPU memory, as it can run with low‑precision quantizations.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| F16 | 0.1 GB | 0.6 GB | Fits | Get file → | |
| Q4_K | 0.2 GB | 0.8 GB | Fits | Get file → | |
| Q8_0 | 0.2 GB | 0.8 GB | Fits | Get file → |
For 8 GB VRAM, the Q4_K build (0.24 GB, 0.8 GB memory at 8k context) is recommended. If more VRAM is available, the F16 build (0.12 GB, 0.6 GB memory) offers higher precision. The Q8_0 build has the same size as Q4_K but higher precision.
What GPU do you need for fastpitch en?
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 Q8_0 (0.2 GB file, about 0.8 GB in use).
Which fastpitch en quant should you choose?
Q4_K (0.2 GB) is the middle option in this repository. The usual sweet spot between file size and quality, and the quant most people start with.
If memory is tight, the smallest file is F16 at 0.1 GB, with the quality trade-offs that come with fewer bits. The unquantized model. It gives original quality and needs the most memory, so it is mainly for conversion or reference.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about fastpitch en
How much VRAM does fastpitch en need?
About 0.8 GB at Q4_K with an 8K context window. That is the 0.2 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 fastpitch en on an 8 GB GPU?
Yes, with Q8_0: the 0.2 GB file needs about 0.8 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of fastpitch en GGUF?
It depends on the quant: F16 is 0.1 GB, Q4_K is 0.2 GB, Q8_0 is 0.2 GB.
Where can I download fastpitch en GGUF?
The files are on Hugging Face in the cstr/fastpitch-en-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 fastpitch en use?
Hugging Face lists the license as cc-by-4.0. 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 FastPitch on 8 GB VRAM?
Yes, the Q4_K build uses 0.8 GB memory at 8k context, which fits within 8 GB VRAM.
Does FastPitch support multiple speakers?
No, the model is single‑speaker only and is trained for a female English voice.
Do I need a separate vocoder?
Yes, the model outputs mel spectrograms; a HiFi‑GAN vocoder is required to convert them to audio.