AI singing platform creates official versions of Chinese-speakers’ voices

2024-04-23

In the rising world of AI, Taiwan has a new AI voice platform. It specializes in creating official versions of Chinese speakers’ voices. A special song featuring the voices of four beloved folk singers has been released to mark the platform going online. Users will be able to put their own compositions and lyrics on the platform, and get them sung by one of these beloved singers within two or three minutes.

Four folk singers collaborated on this new release. But the whole track, from the singers’ voice to the music video, was made by AI.

Li Ming-teh
Singer
I can tell my grandson, “Grandpa sang this song,” and then he’ll say he doesn’t believe it. And I’ll say, “No, really, we recorded it 60 years ago.”

This is the first time these singers have heard their own voices produced by AI.

The singer faces off against her AI rival, showing just how lifelike the imitation is. She says she likes to listen to it.

Fan Yi-wen
Singer
It’s just like my real voice. I don’t usually sing such high ranges, songs in such high keys. So when I heard it, it was very interesting.

Tu Pei-tsen
Singer
It gave me goosebumps. I feel touched. I personally feel it’s very like myvoice’s real timbre.

The driving forces behind the AI music platform are Taiwan AI Labs founder Ethan Tu, and famous campus singer Li Chien-fu.

Li Chien-fu
Singer
It took way less time and money, because, for example, getting a singer into the studio means you have to pay the singer, pay for the studio. All that is a burden on the creator.

Ethan Tu
Taiwan AI Labs founder
In the past, people were always talking about “deepfakes” which were pirate versions and imitations. But we use the same profit-sharing model, and we tidy up the workflows. We’re telling people this could be the future model of the AI voice economy.

Fake voices abound on the internet already. But a platform like this shows that an official AI voice can beat a pirate version. The project gives artists protections and works with a charity for the visually impaired, and gives more creators a platform to make music.

For more Taiwan news, tune in:
Sun to Fri at 9:30 pm on Channel 152
Tue to Sat at 1 am on Channel 53

全球首創! 正版華語AI歌聲平台問世再創民歌盛世

2024-04-23

AI盛行，台灣自力研發的正版華語AI歌聲平台，正式上線，以歌手正版授權為號召，民眾可以將自由創作的歌詞、歌曲，放到平台上，2到3分鐘就會生成一首歌，平台找來知名四位校園民歌手，授權歌聲並親自訓練AI，希望能再創民歌經典。

四位民歌手合作推出新歌，但其實整首歌，從歌聲到MV都是AI做成。

[[歌手李明德]]
“以後可以跟我的孫子講，這首歌曲是爺爺唱的，然後我的孫子會說不相信，(我會說)這真的是爺爺唱的，在六十年前(用AI)錄下來。”

歌手本人通通第一次聽到自己的AI歌聲。全球第一個華語AI歌聲平台，由四位民歌手率先響應，授權歌聲並親自訓練AI，民眾只要將歌曲與歌詞，放到平台上，2到3分鐘後，就能用他們的聲音，生成一首完整的歌曲。

歌手本人與AI現場大PK，歌聲相似度真的極高，就連本人都說讚。

[[歌手范怡文]]
“就是沒有失真。我平常不會唱Range這麼高，就是Key這麼高的歌，所以我聽了也覺得很有趣。”

[[歌手凃佩岑]]
“我真的是起雞皮疙瘩，我覺得有點感動，跟我自己的音質什麼，我自己個人感覺滿像的。”

平台幕後推手是台灣人工智慧實驗室創辦人杜奕瑾，以及知名校園民歌手李建復。

[[民歌教父李建復]]
“時間上少很多，成本上也少很多，因為假如說請一位歌手進錄音室的話，可能有歌手的成本，有錄音室的成本等等，都是對創作者來講是一個負擔。”

[[台灣人工智慧實驗室創辦人杜奕瑾]]
“過去大家一直在講Deepfake、Deepfake，其實都是盜版跟模仿。我們把相關的分潤模式，以及相關的作業流程把它理順之後，我們告訴大家，AI聲音的經濟未來有可能的模式。”

網路AI假聲音氾濫，實驗室喊出以正版打擊盜版，不只保障歌手，也與視障團體公益合作，讓更多人有發揮空間。

更多新聞內容，請鎖定：
民視台灣台（152頻道）週日至週五晚上9:30
民視新聞台（53頻道）週二至週六凌晨1:00