Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -18,6 +18,12 @@ base_model: rednote-hilab/dots.tts-soar
|
|
| 18 |
|
| 19 |
# dots.tts-mf
|
| 20 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
**dots.tts** is a **2B-parameter fully continuous, end-to-end autoregressive (AR) text-to-speech system**. The backbone pairs a semantic encoder, an LLM, and an autoregressive flow-matching acoustic head over a 48 kHz AudioVAE — no discrete codec tokens anywhere in the pipeline.
|
| 22 |
|
| 23 |
This repository hosts **`dots.tts-mf`** — the **CFG-aware MeanFlow distillation** of `dots.tts-soar`. MeanFlow collapses the per-patch ODE to **as few as 2–4 NFE** with a **single model evaluation per step** (CFG is fused into the student — no separate unconditional pass, and `guidance_scale` has no effect at inference time). This is the **recommended checkpoint for low-latency / few-step inference**.
|
|
|
|
| 18 |
|
| 19 |
# dots.tts-mf
|
| 20 |
|
| 21 |
+
<p align="left">
|
| 22 |
+
<a href="https://github.com/rednote-hilab/dots.tts"><img src="https://img.shields.io/badge/GitHub-rednote--hilab%2Fdots.tts-blue?logo=github" alt="GitHub"></a>
|
| 23 |
+
<a href="https://huggingface.co/spaces/rednote-hilab/dots.tts"><img src="https://img.shields.io/badge/%F0%9F%A4%97%20Spaces-Playground-orange" alt="Playground"></a>
|
| 24 |
+
<a href="https://rednote-hilab.github.io/dots.tts-demo/"><img src="https://img.shields.io/badge/Demo%20Page-Live-red" alt="Demo Page"></a>
|
| 25 |
+
</p>
|
| 26 |
+
|
| 27 |
**dots.tts** is a **2B-parameter fully continuous, end-to-end autoregressive (AR) text-to-speech system**. The backbone pairs a semantic encoder, an LLM, and an autoregressive flow-matching acoustic head over a 48 kHz AudioVAE — no discrete codec tokens anywhere in the pipeline.
|
| 28 |
|
| 29 |
This repository hosts **`dots.tts-mf`** — the **CFG-aware MeanFlow distillation** of `dots.tts-soar`. MeanFlow collapses the per-patch ODE to **as few as 2–4 NFE** with a **single model evaluation per step** (CFG is fused into the student — no separate unconditional pass, and `guidance_scale` has no effect at inference time). This is the **recommended checkpoint for low-latency / few-step inference**.
|