As AI technology becomes more and more powerful, there are now many tools that can not only animate static pictures, but also speak and apply your recordings and specify text content. This D-ID can do it, and the operation is super simple. You don’t need to know any technology. You only need to upload the static picture of the person you want, enter or upload the content you want him to say, and you are done. It supports many languages, including Traditional Chinese.
D-ID allows static pictures of people to speak, using your real recordings or transcripts (supports multiple languages)
After entering the D-ID website, there is an explanatory video on the homepage. Anyone who is interested can watch it. Click the Create Video function on the upper right or left:
You will then enter this screen. Choose a presenter to select the character you want. This service provides some foreign faces. You can also upload static pictures yourself, or use the AI generation function on the right (Generate AI presenter):
You need to register an account and provide email or Google or LinkedIn account to log in. Each new account has 20 Credits available:
I used this AI-generated picture of cute and rebellious Ada. After the character is determined, enter the text you want it to speak in the Script field on the right, and Audio will upload the sound file:
Let’s look at the sound part first. After you upload the sound file you want to synthesize with the picture, you will be asked to listen to it. The free version can only be up to 5 minutes in length and the file cannot exceed 10MB. If there is no problem, click GENERATE VIDEO in the upper right corner:
Then this screen will pop up, and the lower right side of the picture will remind you how many Credits will be used. It will be about 20 in 5 minutes, and the free quota will be used up. It is recommended that those who want to try it first, do not use it for so long, it will only take about 30 seconds to 1 minute. No problem. GENERATE:
It will take about 1 to 2 minutes to complete. When completed, your registered email will be notified:
After generating, click on the video:
You can watch and download the video. The generated video cannot be modified. The free version will have a watermark:
The picture below is the result video. D-ID is a foreign service after all. Many Chinese mouth shapes are not correct, and the speed is not very accurate. This part needs to be improved, but I will test the text better (I will share it below). After all, the sound file also needs a speech-to-text recognition, which will have some impact:
The text part is supported by many languages, including Traditional Chinese. Taiwanese is:
Currently, there are only three voices provided in Traditional Chinese, two female and one male, and there are many more voices in English:
Enter what you want to say:
Click the lower left corner to listen:
My paragraph is only 22 seconds long, so it only costs 2 Credits:
As a result, the mouth shape part is obviously better than the sound file:
For those who want more accurate mouth shape matching, it is recommended to use the built-in characters, which will be more natural overall. After the 20 Credits are used up, you must pay for it or register a new account.
Source: KOCPC Chinese