<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: xelta</title>
    <description>The latest articles on DEV Community by xelta (@jhon_wave_fdb4b0e9a0c0a8a).</description>
    <link>https://dev.to/jhon_wave_fdb4b0e9a0c0a8a</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4106292%2Fe6899ff5-27cd-4577-a486-301d36533449.png</url>
      <title>DEV Community: xelta</title>
      <link>https://dev.to/jhon_wave_fdb4b0e9a0c0a8a</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/jhon_wave_fdb4b0e9a0c0a8a"/>
    <language>en</language>
    <item>
      <title>I attempted to make a pipeline that converts into video I'd created myself to an anime. Then I stopped.</title>
      <dc:creator>xelta</dc:creator>
      <pubDate>Wed, 02 Sep 2026 13:36:09 +0000</pubDate>
      <link>https://dev.to/jhon_wave_fdb4b0e9a0c0a8a/i-attempted-to-make-a-pipeline-that-converts-into-video-id-created-myself-to-an-anime-then-i-6p7</link>
      <guid>https://dev.to/jhon_wave_fdb4b0e9a0c0a8a/i-attempted-to-make-a-pipeline-that-converts-into-video-id-created-myself-to-an-anime-then-i-6p7</guid>
      <description>&lt;p&gt;I attempted to make a pipeline that converts into video I'd created myself to an anime. Then I stopped.&lt;br&gt;
So, when you desired to go the anime route with actual footage, you've all had this experience, right? I went with the flow: SD, ControlNet, a lot of extensions and a lot of hope. It was a very easy task to get one clean frame. The nightmarish task of getting the whole frame, without the face jittering, and the lines scrolling from frame to frame was another matter.&lt;br&gt;
The hardest thing when making videos to anime is the anime look. It's the regularity of time. The frames are rendered on one at a time and if the model doesn't see the frames adjacent to it, they will also have the jittery melting effect similar to the frame motion. I spent the entire weekend (and more nodes and settings) working on it.&lt;br&gt;
I attempted Xelta's Video To Anime feature, probably to compare but for the few yester-clips I do it did actually do the tedious work for me. The motion and timing of a video are preserved during upload and download and a video can be downloaded. Video uploading, downloading of anime video and motion and timing of video are preserved.&lt;br&gt;
I'm being generous with my interpretation of the compromise. Having more control is provided by a local pipeline. The model you will be using, the particular LoRA you use, and how hard you can push it, as much as your GPU can do. A hosted tool is one where you sacrifice a degree of control, but you aren't responsible for creating the tool in a day. I'm just pleased to get a fast reel or I want to get a stylized intro, whatever it is. Some things I'd like to make that need a level control on the frame, and I'd definitely take the local stack.&lt;br&gt;
The truthful reporting: If you want an anime version of a clip, but are not willing to be a full time diffusion-pipeline babysitter, it can assist you. If you want to have 100% control of all frames, then use your local setup.&lt;/p&gt;

</description>
    </item>
  </channel>
</rss>
