Thank you for being here. By landing on this page, you are cordially invited to explore and try the new Printto Magicbeat AI voicebanks. We appreciate your interest in our AI vocal synthesis project, and we're excited to have you as part of this journey. Your presence here signifies your curiosity and openness to experience the possibilities of AI-generated voices. We sincerely thank you for your participation and look forward to your engagement with these innovative voicebanks.
We are thrilled to announce the release of his brand-new Thai DiffSinger voicebank.
We've come so far to the first ever Thai AI voicebank!
An AI voice library for OpenUTAU supports 3 languages; Thai, English, and Japanese in 1 voice library.
While this voice library accommodates multiple languages, it is still recommended to use Native voice libraries for the most accurate pronunciation and phoneme timing.
Requires OpenUTAU v0.1.501 (Beta) or more.
This voice library utilizes the publically available Singing Dataset "TIGER" and "PJS", meant to be used for parallel training in SVS Synthesizers.
The first and latest version of Printto Magicbeat Thai AI with Thai characters fully compatible.
The V2 voice library supports OpenUTAU's "tension" curve adjustment.
Requires OpenUTAU v0.1.501 (Beta) or more.
Male English AI voice library for OpenUTAU with mellow/soft vocal.
This voice library utilizes the publically available Singing Dataset "TIGER", meant to be used for parallel training in SVS Synthesizers.
Male Japanese AI voice library for OpenUTAU with mellow/soft vocal.
This voice library utilizes the publically available Singing Dataset "PJS".
An AI voice library for OpenUTAU trained with Printto Magicbeat's UTAU voice samples.
This voice library utilizes the publically available Singing Dataset "TIGER", meant to be used for parallel training in SVS Synthesizers.
Male Japanese AI voice library for OpenUTAU ENUNU.
Diff-SVC is an open-source AI voice conversion software developed by a team of CUHK university students for research purposes. It allows for the creation of new voice models based on user input data.
Trace and emulate the vocal and singing styles of other individuals with remarkable accuracy using this cutting-edge model.
Create cover songs or replicate specific vocal characteristics with the flexibility and precision offered by this model.
The model is based on 1 hour of Thai and English singing datasets and trained for 210,000 steps.
To make it easier to manage the policy, the download link will be sent to your email provided below.
Please ensure that you read the README.txt file provided with the downloaded voicebanks before using them. It contains important instructions and guidelines that must be followed.