From c96741137380e40c4b3cf2424c84f473edebe777 Mon Sep 17 00:00:00 2001
From: cantabile-kwok <58417810+cantabile-kwok@users.noreply.github.com>
Date: Sun, 8 Oct 2023 23:45:33 +0800
Subject: [PATCH] finish vocoder

---
 README.md         | 2 +-
 hifigan/README.md | 6 ++++--
 2 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/README.md b/README.md
index 7cb6ebc..1c06029 100644
--- a/README.md
+++ b/README.md
@@ -1,4 +1,4 @@
-# \[Working in Progress\] VoiceFlow: Efficient Text-to-Speech with Rectified Flow Matching
+# VoiceFlow: Efficient Text-to-Speech with Rectified Flow Matching
 > This is the official implementation of [VoiceFlow](https://arxiv.org/abs/2309.05027).
 
 ![traj](resources/traj.png)
diff --git a/hifigan/README.md b/hifigan/README.md
index 0888738..e85d09c 100644
--- a/hifigan/README.md
+++ b/hifigan/README.md
@@ -8,11 +8,13 @@ The detailed information is:
 | LJSpeech | 16k           | 256      | 1024          | True | 
 | LibriTTS | 16k           | 200      | 800           | True |
 
-The trained checkpoint on both datasets are provided online. You can unzip them to sub-folders in `exp/`.
+The trained checkpoint on both datasets are provided online. You can unzip them to sub-folders in `exp/`:
+* LJSpeech: [link](https://huggingface.co/cantabile-kwok/hifigan-ljspeech-1024-256/resolve/main/train_hifigan.ljspeech.zip)
+* LibriTTS: [link](https://huggingface.co/cantabile-kwok/hifigan-libritts-800-200/resolve/main/train_hifigan.libritts.zip)
 
 Vocoding can be done by 
 ```shell
 cd ../; source path.sh; cd -;  # if path.sh not activated
-bash generation.sh --dataset "ljspeech/libritts" --eval_dir /path/that/contains/feats.scp
+bash generation.sh --dataset "ljspeech or libritts" --eval_dir /path/that/contains/feats.scp
 ```
 The program will read feats.scp in $eval_dir and synthesize audio to save in that dir.