【Hacker News搬运】通过将眼睛转向看相机来纠正视频中的眼神交流

hackernews

Title: Correct eye contact in videos by redirecting the eyes to look at the camera

通过将眼睛转向看相机来纠正视频中的眼神交流

Text:

Url: https://www.sievedata.com/functions/sieve/eye-contact-correction

很抱歉，我无法直接访问互联网来查看或分析链接内容。不过，我可以提供一种方法来使用JinaReader或其他类似的文本分析工具来抓取和总结网页内容，包括翻译非中文内容的过程。

以下是一个基本的步骤指南，说明如何使用Python和JinaReader（假设它是一个可以用于文本处理的Python库）来抓取网页内容、进行文本分析以及翻译非中文内容：

1. **安装必要的库**：
   - `requests`：用于发送HTTP请求。
   - `BeautifulSoup`：用于解析HTML。
   - `jina_reader`：用于文本分析。
   - `googletrans`：用于翻译文本（这是一个常用的翻译库）。

```bash
pip install requests beautifulsoup4 jina-reader googletrans==4.0.0-rc1

编写Python脚本：

import requests
from bs4 import BeautifulSoup
from jina_reader import JinaReader
from googletrans import Translator, LANGUAGES

# 网页URL
url = "https://www.sievedata.com/functions/sieve/eye-contact-correction"

# 发送请求并获取响应
response = requests.get(url)
soup = BeautifulSoup(response.text, 'html.parser')

# 提取文本内容
text_content = soup.get_text()

# 使用JinaReader分析文本
jina_reader = JinaReader()
summary = jina_reader.summarize(text_content)

# 翻译文本（假设内容不是中文）
translator = Translator()
if 'zh-cn' not in LANGUAGES.values():
    summary = translator.translate(summary, dest='zh-cn').text

# 输出总结
print(summary)

运行脚本：
- 将上述脚本保存为一个.py文件。
- 在命令行中运行脚本，例如：python script.py。

请注意，上述脚本只是一个示例，实际使用中可能需要根据网页的具体结构和内容进行调整。例如，JinaReader可能没有提供一个summarize方法，或者你需要根据网页的实际内容提取文本。

此外，翻译功能使用的是googletrans库，它依赖于Google Translate API，因此可能需要处理API的请求限制和错误。

由于我无法访问互联网来执行上述操作，因此无法提供一个实际的输出结果。如果你需要处理非中文内容，确保在脚本中添加了适当的错误处理和异常管理。

        
## Post by: thunderbong
        
### Comments: 
        
**albert_e**: Practically --<p>I feel hardware technology can improve further to allow under-the-LED-display cameras .... so that we can actually look at both the camera and the screen at the same time.<p>(There are fingerprint sensors under mobile screens now ...and I think even some front facing cameras are being built in without sacrificing a punch hole &#x2F; pixels. There is scope to make this better and seamless so we can have multiple cameras if we want behind a typical laptop screen or desktop monitor.)<p>This would make for a genuine look-at-the-camera video whether we are looking at other attendees in a meeting or reading off our slide notes (teleprompter style).<p>There would be no need to fake it.<p>More philosophically --<p>I don&#x27;t quite like the normalization of AI tampering with actual videos and photos casually -- on mobile phone cameras or elsewhere. Cameras are supposed to capture reality by default. I know there is already heavy noise reduction, color correction, auto exposure etc ... but no need to use that to justify more tampering with individual facial features and expressions.<p>Videos are and will be used for recording humans as they are.  The capturing of their genuine features and expressions should be valued more. Video should help people bond as people with as genuine body lanuage as possible. Videos will be used as memories of people bygone. Videos will be used as forensic or crime scene evidence.<p>Let us protect the current state of video capture. All AI enhancements should be marketed separately under a different name, not silently added into existing cameras.
> **albert_e**: 实际上——<p>我觉得硬件技术可以进一步改进，以允许LED显示屏下的摄像头。。。。这样我们就可以同时看到相机和屏幕<p> （现在移动屏幕下有指纹传感器……我认为即使是一些前置摄像头也在不牺牲打孔像素的情况下被内置。如果我们想在一个典型的笔记本电脑屏幕或台式电脑显示器后面安装多个摄像头，我们还可以让它变得更好、无缝。）<p>无论我们是在看会议中的其他与会者还是在读幻灯片（提词器风格），这都有助于真实地观看摄像头视频<p> 没有必要伪造它。<p>更哲学地说——<p>我不这么认为；我不太喜欢人工智能随意篡改实际视频和照片的正常化——在手机摄像头或其他地方。默认情况下，相机应该捕捉现实。我知道已经有了大量的降噪、色彩校正、自动曝光等功能。。。但无需以此为理由，进一步篡改个人面部特征和表情<p> 视频现在和将来都将用于记录人类的真实状态。应该更加重视捕捉他们的真实特征和表情。视频应该帮助人们尽可能地用真实的肢体语言建立联系。视频将成为人们逝去的回忆。视频将被用作法医或犯罪现场的证据<p> 让我们保护视频捕获的当前状态。所有AI增强功能都应该以不同的名称单独销售，而不是默默地添加到现有的相机中。
            
**Retr0id**: Does what it says on the tin, but honestly I find the &quot;uncorrected&quot; video more comfortable to watch.
> **Retr0id**: 照锡上写的做，但老实说，我发现&quot；未纠正&quot；视频观看更舒适。
            
**xnx**: Nvidia has free Broadcast software with an eye contact feature: <a href="https:&#x2F;&#x2F;www.nvidia.com&#x2F;en-us&#x2F;geforce&#x2F;news&#x2F;jan-2023-nvidia-broadcast-update&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.nvidia.com&#x2F;en-us&#x2F;geforce&#x2F;news&#x2F;jan-2023-nvidia-br...</a><p>It&#x27;s from January 2023, so I don&#x27;t know if they&#x27;ve improved it further since then.<p>The video conferencing software providers have been way to slow to put whoever is speaking top-center (near where the camera typically is).
> **xnx**: Nvidia提供具有眼神交流功能的免费广播软件：<a href=“https:”www.Nvidia.com“en us”geforce“news”2021年1月Nvidia-Broadcast-update“rel=”nofollow“>https:”&#x2F；www.nvidia.com；en us；geforce；news；2023年1月，nvidia-br…</a><p>；从2023年1月开始，所以我不知道；不知道他们是否；从那以后，我进一步改进了它<p> 视频会议软件提供商一直在缓慢地将发言的人放在最中心（靠近摄像机通常所在的位置）。
            
**richdougherty**: Kudos to the dev for coming up with the eye position fixing solution.<p>Building further on this idea, I wonder if instead of changing the image to look at the camera, we could change the &quot;camera&quot; to be where we&#x27;re looking.<p>In other words we could simulate a virtual camera somewhere in the screen, perhaps over the eyes of the person talking.<p>We could simulate a virtual camera by using the image of the real camera (or cameras), constructing a 3D image of ourselves and re-rendering it from the virtual camera location.<p>I think this would be really cool. It would be like there was a camera in the centre of our screen. We could stop worrying about looking at the camera and look at the person talking.<p>Of course this is all very tricky, but does feel possible right now. I think the Apple Vision Pro might do something similar already?
> **richdougherty**: 开发人员提出了眼睛位置固定解决方案，值得称赞<p> 在这一想法的基础上，我想知道我们是否可以改变&quot；相机”；成为我们；重新审视<p> 换句话说，我们可以在屏幕的某个地方模拟一个虚拟摄像头，也许在说话的人的眼睛上方<p> 我们可以通过使用真实相机（或多个相机）的图像来模拟虚拟相机，构建我们自己的3D图像，并从虚拟相机位置重新渲染<p> 我认为这真的很酷。这就像我们屏幕中央有一个摄像头。我们可以不用担心看着相机，看着说话的人<p> 当然，这一切都很棘手，但现在确实有可能。我认为Apple Vision Pro可能已经做了类似的事情？
            
**not_a_bot_4sho**: I&#x27;ve never seen an implementation of this that wasn&#x27;t super creepy past the initial tech demo
> **not_a_bot_4sho**: 我；我从未见过这样的实现方式；在最初的技术演示之后，这并不可怕