Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Official repository of DailyTalk: Spoken Dialogue Dataset for Conversational Text-to-Speech, ICASSP 2023
| Date | Stars |
|---|---|
| 2026-07-24 | 260 |
| 2026-07-25 | 260 |
| 2026-07-28 | 260 |
| 2026-07-30 | 260 |
| 2026-07-31 | 259 |
| 2026-08-06 | 259 |
Today
— stars today
This week
-1 stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
[](https://arxiv.org/abs/2207.01063)
# DailyTalk: Spoken Dialogue Dataset for Conversational Text-to-Speech
### Keon Lee<sup>\*</sup>, Kyumin Park<sup>\*</sup>, Daeyoung Kim
In our [paper](https://arxiv.org/abs/2207.01063), we introduce DailyTalk, a high-quality conversational speech dataset designed for Text-to-Speech.
<p align="center">
<img src="img/dailytalk_table.png" width="70%">
</p>
<p align="center">
<img src="img/dailytalk_model.png" width="70%">
</p>
**Abstract:** The majority of current Text-to-Speech (TTS) datasets, which are collections of individual utterances, contain few conversational aspects. In this paper, we introduce DailyTalk, a high-quality conversational speech dataset designed for conversational TTS. We sampled, modified, and recorded 2,541 dialogues from the open-domain dialogue dataset DailyDialog inheriting its annotated attributes. On top of our dataset, we extend prior work as our baseline, where a non-autoregressive TTS is conditioned on historical information in a dialogue. From the baseline experiment with both general and our novel metrics, we show that DailyTalk can be used as a general TTS dataset, and more than that, our baseline can represent contextual information from DailyTalk. The DailyTalk dataset and baseline code are freely available for academic use with CC-BY-SA 4.0 license.
# Dataset
You can download our [dataset](https://drive.google.com/drive/folders/1WRt-EprWs-2rmYxoWYT9_13omlhDHcaL). Please refer to [Statistic Details](https://github.com/keonlee9420/DailyTalk/tree/26239109fba7064a60ee02e5d4846152648142df#statistic-details) for details.
# Pretrained Models
You can download our [pretrained models](https://drive.google.com/drive/folders/1RrmWzJM1iWhQg2_wvDHlszVDJRV60Wfu?usp=sharing). There are two different directories: 'history_none' and 'history_guo'. The former has no historical encodings so that it is not a conversational context-aware model. The latter has historical encodings following [Conversational End-to-End TTS for Voice Agent](https://arxiv.org/abs/2005.10438) (Guo et al., 2020).
Toggle the type of history encodings by
```yaml
# In the model.yaml
history_encoder:
type: "Guo" # ["none", "Guo"]
```
# Quickstart
## Dependencies
You can install the Python dependencies with
```
pip3 install -r requirements.txt
```
Also, `Dockerfile` is provided for `Docker` users.
## Inference
You have to download both [our dataset](https://drive.google.com/drive/folders/1WRt-EprWs-2rmYxoWYT9_13omlhDHcaL). Download [pretrained models](https://drive.google.com/drive/folders/1RrmWzJM1iWhQg2_wvDHlszVDJRV60Wfu?usp=sharing) and put them in `output/ckpt/DailyTalk/`. Also unzip `generator_LJSpeech.pth.tar` or `generator_universal.pth.tar` in hifigan folder. The models are trained with unsupervised duration modeling under transformer building block and the history encoding types.
Only the batch inference is supported as the generation of a turn may need contextual history of the conversation. Try
```
python3 synthesize.py --source preprocessed_data/DailyTalk/val_*.txt --restore_step RESTORE_STEP --mode batch --dataset DailyTalk
```
to synthesize all utterances in `preprocessed_data/DailyTalk/val_*.txt`.
# Training
## Preprocessing
- For a **multi-speaker TTS** with external speaker embedder, download [ResCNN Softmax+Triplet pretrained model](https://drive.google.com/file/d/1F9NvdrarWZNktdX9KlRYWWHDwRkip_aP) of [philipperemy's DeepSpeaker](https://github.com/philipperemy/deep-speaker) for the speaker embedding and locate it in `./deepspeaker/pretrained_models/`. Please note that our pretrained models are not trained with this (they are trained with `speaker_embedder: "none"`).
- Run
```
python3 prepare_align.py --dataset DailyTalk
```
for some preparations.
For the forced alignment, [Montreal Forced Aligner](https://montreal-forced-aligner.readthedocs.io/en/latest/) (MFA) is used tExcerpt of 7,282 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:d59a5aa09a5ed553, topic:text-to-speech, topic:tts, topic:speech-synthesis