DO

dots tts soar

DotsGGUF by cstrUpdated 2026-08-02
Download on Hugging Face
Parameters2.2B
Context—
Layers—
Architecturedots-tts-spk
Smallest quant0.0 GB
Quick answer

dots tts soar needs about 0.9 GB of memory at F16 (a 0.4 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 7 quants, from 0.0 GB (F16) to 4.6 GB (F16).

dots tts soar – a 2.2B‑parameter continuous‑latent text‑to‑speech model

dots tts soar is a 2.2B‑parameter autoregressive TTS model. It generates 48 kHz speech in a continuous latent space and is designed for use with CrispASR. The model is released under the Apache‑2.0 license.

Good at

  • High‑fidelity 48 kHz speech generation
  • Continuous latent synthesis without a discrete codec
  • Compatible with CPU inference via ggml

Limits

  • Requires a separate vocoder companion file
  • DiT head must remain in F16 precision
  • No vision or multimodal support

Best for

  • Embedded TTS on devices with ggml support
  • Batch speech generation for text‑based applications
  • Low‑latency voice cloning with optional speaker encoder

Who it’s for. Developers building lightweight TTS engines who need high‑quality 48 kHz output without Python or PyTorch dependencies. Ideal for C/C++ projects that use CrispASR and ggml.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
F160.0 GB 0.5 GB
Fits Get file →
Q4_K0.3 GB 0.9 GB
Fits Get file →
Q8_00.3 GB 0.9 GB
Fits Get file →
F160.4 GB 0.9 GB
Fits Get file →
Q4_K2.3 GB 3.1 GB
Fits Get file →
Q8_03.1 GB 4.0 GB
Fits Get file →
F164.6 GB 5.7 GB
Fits Get file →

For 8 GB VRAM, use the Q4_K quant with file size 0.34 GB (est. 0.9 GB memory). For 4 GB VRAM, the Q4_K quant at 2.32 GB (est. 3.1 GB memory) is suitable. If more memory is available, the F16 quant at 4.64 GB (est. 5.7 GB memory) offers the highest precision.

What GPU do you need for dots tts soar?

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 (3.1 GB file, about 4.0 GB in use).
  • 6 GB VRAM: the best quant that fits is F16 (4.6 GB file, about 5.7 GB in use).

Which dots tts soar quant should you choose?

F16 (0.4 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 F16 at 0.0 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

Architecturedots-tts-spk
Parameters2.2B

Source

Quantized bycstr
FamilyDots
Licenseapache-2.0
Downloads (HF)2,213
Last updated2026-08-02

Frequently asked questions about dots tts soar

How much VRAM does dots tts soar need?

About 0.9 GB at F16 with an 8K context window. That is the 0.4 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 dots tts soar on an 8 GB GPU?

Yes, with F16: the 4.6 GB file needs about 5.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of dots tts soar GGUF?

It depends on the quant: F16 is 0.0 GB, Q4_K is 0.3 GB, Q8_0 is 0.3 GB, F16 is 0.4 GB, Q4_K is 2.3 GB, Q8_0 is 3.1 GB and 1 more.

Where can I download dots tts soar GGUF?

The files are on Hugging Face in the cstr/dots-tts-soar-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 dots tts soar 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 dots tts soar on a CPU without GPU acceleration?

Yes. The model runs on CPU via ggml, with optional Metal, CUDA, or Vulkan acceleration for faster inference.

Do I need to re‑quantize the DiT head for better performance?

No. The DiT head is kept at F16 in every quant; re‑quantizing it degrades audio quality.

Is a vocoder required for inference?

Yes. Each core GGUF needs the vocoder companion; the vocoder is auto‑discovered next to the core file.

Similar models

Browse all →

More from cstr

All by cstr →

Other tiny models

All tiny models →