Speech AI Laboratory
Information about "Speech AI Research Center (SARC)" will be updated at https://sites.google.com/nycu.edu.tw/sarc !
This page will only update the relevant information of "Speech Lab"!
此網頁將只更新“語音實驗室”的相關資訊!
vision
- Long-Term: Universal Human-Machine Interface
It's conceivable now that in the near future, humans will be able to use spoken language as a direct means of communicating with machines and directing them to work together. This means that computers will need to be able to accurately interpret human language and have a deep understanding of how the world works, as well as the ability to use a range of tools - including programming - to carry out human commands.
- Short-Term: Multi-Modal Foundation Model
We want to develop a robot that can independently acquire knowledge of the human language and a basic understanding of the world solely by watching television.
News
- 「2023客語AI應用黑客松·創意發想大賽」及「2023客語語音辨認競賽」
- 總統府記者會直播字幕(使用我們的即時字幕系統)
執政七年重新定義台灣讓世界重新看見台灣記者會https://www.youtube.com/live/lIm7faMXA0M
- Taiwanese/Hakka Speech-enabled ChatGPT
公視台語台新聞:ChatGPT教臺語? 學者當咧開發AI程式拚傳承 [YouTube]
換日線Crossing:國際趨勢/科技/阿善 Café 的世界分館 2023/05/05:【台大演講筆記】ChatGPT 會說台語?──台灣團隊研究 AI,為本土語言復振努力|換日線 Crossing
Our Working Horses
- 6 GPU servers
GPU1: 1080*8, 128GB
GPU2: 1080ti*8, 256GB
GPU3: 2080ti*10, 256GB
GPU4: 2080ti*8, 256GB
GPU5: 3090*10, 256GB
GPU6: 3090*10, 256GB
- 2 NASs
QNAP 1635ax * 16 slots ~ 100TB
Synology RS4021xs+ * 16 slots ~ 100TB
- > 20 PC+GPUs
Hakka Speech Corpus
Target: 5 dialects
each dialect will collect 150 speakers, 300 hours speech for ASR, 2 speakers, 60 hours speech for TTS
Status
Sixian finish
Hailu ongoing
- News
WiTMed (smart medicine)
- Handfree
- Mobility
- Standardization
- Flexibility
Formosa Speech in the Wild
- Project Homepage
- Hokkien <--> English Translation
- TAT & TAT_S2ST_Benchmark Corpus
- Formosa TV News, 2020/12/17
- Hakka TV News, 2022/02/21
3C裝置講客聽得懂 產官學共建客語語音庫【客家新聞20220221】 [Youtube (Hakka)]
Automatic Media Subtitle Generation
- Real-Life Examples of Our Subtitling System
PS: Turn on Chinese subtitles by pressing the CC icon
Chinese Text to TaiwanesE/Hakka Speech Synthesis
- Chinese Text to Taiwanese Speech Synthesis
- Chinese Text to Hakka Speech Synthesis
Voice Bank For All
- Voice Banking Service
- Personalized Speech Synthesis
President Tsai’s Personalized TTS (please save as pptx and play)
- Speech-to-Speech Voice Conversion
- Movie Dubbing
Untold HerStory (流麻溝十五號) - Chiang Ching-Kuo's Speech