CSTR-Edinburgh / merlin

This is now the official location of the Merlin project.
http://www.cstr.ed.ac.uk/projects/merlin/
Apache License 2.0
1.31k stars 442 forks source link

Create new language #105

Open dhm42 opened 7 years ago

dhm42 commented 7 years ago

Hi, I want to use Merlin with French langage. Can you guide me throught the process of creating a new langage (and voice).This will be helpfull for people who want to participate in Merlin developpement. I have a corpus of wav files with their text transcription. What does Merlin need as data (apart from alignment files and audio files). What are the tools needed to do the linguistic and lexical analysis of the text files and what do they generate. Is there any lexical or linguistic features that needs to be developed for French Langage? Any help will be appreciated, thanks in advance for your help.

dreamk73 commented 7 years ago

You need a linguistic frontend to process the text to get relevant linguistic features for Merlin. You can start with a very simple set of features defined in the question file and see how it goes from there. Typically if you start from scratch, I would see if there is any French voice available for Festival and use that. At the very least you need information about each phoneme, accents, phrase boundaries, and counters for how many phonemes there are in the syllable, how many syllables in the phrase, etc (counting both forward and backward).

If you can't find it, I would write a script from scratch using the input transcriptions and having a small number of function words which never receive an accent and use commas in the sentences to denote phrase boundaries.

shartoo commented 7 years ago

@dreamk73 i want to use Merlin with Chinese language ,which is same as French when constructing from scratch.There is Chinese voice dataset like 'THCHS30'.Please share an example or tutorial .

Jackiexiao commented 7 years ago

@shartoo I want to use Merlin with Chinese language too. Do you have any idea now?

shartoo commented 7 years ago

@Jackiexiao not yet.I'm trying,but not focusing on this topic.I have to do image processing work. You can keep my contact QQ:604135528 or gmail: shartoo518@gmail.com

Jackiexiao commented 7 years ago

@dhm42 I strongly recommend the tutorial from Columbia University. It's the best tutorial for speech synthesis I have ever seen ( for a new beginner) !

chazo1994 commented 7 years ago

@Jackiexiao I cannot access this link Merlin Instructions and Troubleshooting in your tutorial.

Jackiexiao commented 7 years ago

@chazo1994 sorry, only columbia student can access it

ecooper7 commented 7 years ago

we've just updated that link to be public.

On Sat, Aug 19, 2017 at 3:23 AM, 鉴津Jackie notifications@github.com wrote:

@chazo1994 https://github.com/chazo1994 sorry, only columbia student can access it

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/CSTR-Edinburgh/merlin/issues/105#issuecomment-323506587, or mute the thread https://github.com/notifications/unsubscribe-auth/ABReNkmeni_CUCv4HmCS4L2NLDBt-3L5ks5sZo1igaJpZM4MRCs4 .