Better Vietnamese voices support for Piperok

[duoc-van]

Hello Piper community,

I would like to request better Vietnamese voice support for Piper.

Piper is already a very useful tool for subtitle work, voice testing, and narration workflow because it is lightweight, fast, local, and easy to test line by line. This makes it especially valuable for users who need to check pauses, punctuation, and reading rhythm carefully before generating a full result.

At the moment, the biggest missing piece is Vietnamese voice quality and variety.

If Piper could have just 2 or 3 better Vietnamese voices, it would become much more useful for practical real-world work such as subtitle dubbing, narration, and documentary-style voiceover. For example:

one natural male voice,
one natural female voice,
and possibly one voice suitable for documentary narration.

Many users do not need a very complex system. What they really need is:

more natural Vietnamese pronunciation,
more stable reading rhythm,
better overall voice quality,
and the ability to test short lines quickly before full use.

This is especially important for subtitle and narration workflows, where users often need to test a single sentence, a proper name, or a punctuation pattern before processing a full file. Piper already does this workflow very well. Better Vietnamese voices would make the system much more complete.

Vietnamese is a large and active user base, and stronger Vietnamese voice support would make Piper much more valuable for practical everyday use.

Thank you for building and maintaining Piper.

[phineas-pta]

https://github.com/nghimestudio/nghitts

[ken107]

These voices are quite good. Would love to have them integrated into Piper.

[phineas-pta]

piper chạy dc file onnx mà

[ken107]

yea, i mean, thêm vào voicelist chính thức cuả piper trên huggingface để nó tự động púsh ra cho mấy user của mình khỏi phải viet thêm code. Tuy nhiên hôm qua đọc code của nghitts thấy có làm khá nhiều preprocessing. Đang nghĩ coi nhờ AI làm sao cho mình đỡ tốn công kaka.

[phineas-pta]

https://github.com/nghimestudio/vietnormalizer

[ken107]

https://github.com/Read-Aloud/nghitts/tree/read-aloud
In the process of adding support for nghitts in the Read Aloud browser extension. The quality is quite amazing really. I love the Nguyen Ngoc Ngan voice for reading stories.

That text processing logic is good. Vietnamese does need this special handling. Just one thing remaining is populating the acronyms.csv file with the most common abbreviations. Unfortunately we have a lot of acronyms and I can't find any definitive list online. Thinking about using a webservice that uses AI to expand an acronym, and build a database of acronyms as we encounter them.

Edit: https://blog.readaloud.app/2026/07/read-aloud-co-them-giong-oc-tieng-viet.html

[duoc-van]

Hi, I would like to use these .onnx voice models with Piper / Subtitle Edit.

I can only find the .onnx files, but I do not see the matching .onnx.json config files.

Could you please upload the corresponding .onnx.json files for each voice model?
For example:
banmai.onnx
banmai.onnx.json

If these models are not directly compatible with Piper, could you please provide instructions or a sample command to run them with Piper?

My goal is to load the voice directly in Subtitle Edit → Text-to-Speech (Piper).
Thank you.

[phineas-pta]

trong link có file json mà

[duoc-van]

Cám ơn nhiều, đã hổ trợ chia sẽ, lần trước lấy về vào trang này - nên chỉ có như hình :

[phineas-pta]

à do mấy file json giống nhau uết nên mình gộp chung lại th6i