Hello guys, I have been working on this project for around 6 months now. Even though it started a year later than DeepL Japanese translation (March 2020), the translation quality has surpassed DeepL in a few benchmarks and there are still ways to further improve it.
You can refer to this reddit post for more details: https://www.reddit.com/r/visualnovels/comments/qnsgd2/from_sugoi_translator_developer_today_i_will_send/
The most straightforward and effective strategy is to train the model with more in-domain data. However getting the right data can be time consuming and expensive.
My goal is to get around at least 1 million lines which cost around 1700 USD. Every 5 dollars equals 3000 lines to the training data.
Feel free to donate the amount you want to contribute. 5 dollars, 10 dollars, 50 dollars, or even 100 dollars and more. Each 5 dollars MATTERS.
If the goal is not reached, I will spend my own money to cover the rest of the amount for at least 1 million lines.
Working with me are a few very passionate people who also happened to have free time to help out. I think from now until end of December 2021 is the last time we could have such a team to dedicate to this cause. The faster we reached the goal, the faster we can start getting data and process it.
I hope this effort will let us finally have an offline model that is better than DeepL in visual novel and manga translation
LET'S SHOW DEEPL WHAT WE GOT
PS: for any future donation, if you are in my discord channel. Leave your name if you wanted, and I'll thank you personally
