Downloads · 30 days
4
22% of all-time downloads
thorsten_tts_train_tacotron2_raw_phn_espeak_ng_german
jjiang4/thorsten_tts_train_tacotron2_raw_phn_espeak_ng_german
thorsten_tts_train_tacotron2_raw_phn_espeak_ng_german is a text-to-speech model from jjiang4. Use it when you need text read aloud. It is set up for espnet. The card lists the license as cc-by-4.0.
Tacotron 2 trained on the Thorsten-Voice Dataset 2022.10 German single male speaker corpus, using the egs2/thorsten/tts1 recipe in espnet.
All-time downloads
18
Public
Repo size
215 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pth107 MB · 100%
At a glance
- Task
- Text-to-Speech
- Library
- espnet
- License
- cc-by-4.0
- Access
- Public
- Created
- Aug 27, 2026
- Updated
- Aug 27, 2026
- SHA
- ba352e56
Datasets
Model card
From the Hugging Face model README
ESPnet2 TTS model
jjiang4/thorsten_tts_train_tacotron2_raw_phn_espeak_ng_german
Tacotron 2 trained on the Thorsten-Voice Dataset
2022.10 German single male speaker corpus,
using the egs2/thorsten/tts1 recipe in
espnet.
Griffin-Lim is the default vocoder. For much better quality, pair this model
with jjiang4/thorsten_hifigan_ft_ljspeech.
Demo: How to use in ESPnet2
cd espnet/egs2/thorsten/tts1
./run.sh --skip_data_prep false --skip_train true \
--download_model jjiang4/thorsten_tts_train_tacotron2_raw_phn_espeak_ng_german
from espnet2.bin.tts_inference import Text2Speech
tts = Text2Speech.from_pretrained(
"jjiang4/thorsten_tts_train_tacotron2_raw_phn_espeak_ng_german"
)
wav = tts("im prozess wurden aber nur vierzig fälle thematisiert.")["wav"]
Results
100-utterance test set. See the recipe README for how these were measured.
| System | MCD | log-F0 RMSE | UTMOS | WER (%) | CER (%) |
|---|---|---|---|---|---|
| Ground truth | - | - | 3.29 ± 0.21 | 6.2 | 3.2 |
| Tacotron 2 + Griffin-Lim | 10.27 ± 1.07 | 0.256 ± 0.056 | 1.42 ± 0.13 | 10.0 | 4.4 |
| Tacotron 2 + HiFi-GAN | 5.83 ± 1.36 | 0.268 ± 0.061 | 3.04 ± 0.28 | 10.2 | 4.1 |
Environments
- espnet version:
espnet 202604 - pytorch version:
pytorch 2.8.0+cu128 - python version:
3.10.14
- Task
- Text-to-Speech
- Library
- espnet
- License
- cc-by-4.0
- Languages
- de
- Created
- Aug 27, 2026
- Updated
- Aug 27, 2026