336
请先选择匹配项
请先选择片中职务
分享到微信朋友圈
1.BACKGROUNDPROJECT
Ambrosia strongly supports the Chinese team in the Asian Games, hoping that the Chinese e-sports team can show their talents in the competition, and in such a grand sports meeting, of course, the most advanced technology should be used in advertising.
安慕希大力支持亚运会中国队伍，希望中国电竞队伍能够在大赛上展露头角，这样盛大的运动会上，广告当然也要用最先进的技术。
Isn't that a coincidence? Angolis keeps pace with the times and is in the business of AI drawing and AI animation, so we hit it off and teamed up to produce this promotional animation generated by AI training.
这不巧了吗？安戈力与时俱进，正在开展AI绘图、AI制作动画的业务，于是我们一拍即合，联手制作了这支由AI训练生成的宣传动画。
This creation is based on the theme of various peripherals of the participating teams, so that posters and keychains can also speak, which sounds quite crazy, let's see the actual effect together!
这次的创作以参赛队伍的各种周边为题，让海报、钥匙扣也能说话，听起来挺瘆人的，让我们一起看看实际的效果吧！
接到任务的第一步，就是解决“说话”的问题，既然是AI作品，当然不能用选手的原声配音了，这时候AI的创造力就派上了用场。
We intercepted the contestants' various interview sources, trained the contestants' voices with the FastVits tool, and after hours of training, we got the contestants' timbre and even the accents were imitated in place.
我们截取了选手的各类采访音源，将选手的声音使用FastVits工具训练，经过数个小时的训练，我们得到了选手的音色，甚至口音也会被模仿到位。
Although the AI imitates it very similarly, after experiments, it was found that the voice output by the AI still lacks some emotion, so we recorded the dry lines with feelings, and then used the trained voice model of the player to output, so that we got the lines with the player's timbre and accent!
虽然AI模仿的已经非常像了，但是经过实验发现AI输出的声音还缺少了一些情感，于是我们录制了带感情的台词干声，再使用训练好的选手声音模型进行输出，这样我们就得到了具备选手音色和口音的台词啦！
The next step is to train the mouth shape according to the lines, take the player's avatar, use Sadtalker, D-ID tools to train the mouth shape, after a whole day of training, a static picture will move the mouth to speak, smooth and natural.
下一步就是根据台词训练口型了，截取选手的头像，使用Sadtalker、D-ID工具训练口型，经过一整天的训练，一张静态的图片就动嘴说起话来，流畅又自然。
Finally, the animation is synthesized, and the player's avatar with a transparent channel is spliced to the original poster and peripheral materials, and the amazing effect of static pictures is achieved!
最后就是动画合成了，用带透明通道的选手头像拼接到原本的海报和周边素材上，就达到了静态图片说话的惊人效果！
3.CHARACTER SETTING
The image design produced this time was provided by the customer, and the blue and white Ambrosia yogurt echoed the team uniform, and our illustrator's top cutout technology was used to separate the images one by one, so as to facilitate the training of AI.
这次制作的形象设计由客户提供，蓝白配色的安慕希酸奶与队服相呼应，动用我们插画师的顶级抠图技术把一个个形象都分离开，这样才方便AI的训练工作。
4.CAMERA VOICE DESIGN
The story chooses several different scenes, the posters held up by fans at the competition, the unfinished posters and standing signs in the office, the keychains hanging on the fans' schoolbags on the subway, and the large posters above the vending cabinets in the subway.
故事选取了几个不同的场景，大赛上粉丝们举起的海报，办公室里还没有完成的海报和立牌，地铁上粉丝书包上挂着的钥匙扣，地铁内售货柜上方的大海报，台词十分接地气，好像真正的选手在现实中说话，调侃着自己接的广告，打趣又自然。
内容由作者原创，转载请注明来源，附以原文链接
https://open.6pian.cn/xuanchuan/257029.html全部评论
发布者暂未开放下载权限
我们正努力使更多创作人开放下载权限，敬请期待...
确定下载本片吗?
提交成功
待作品发布者确认
无法认领此作品
只有实名认证的用户才可以进行作品创作人关联哦~
提交成功
待创作人确认
取消邀请
表情
添加图片
发表评论