<?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: Alex Hu</title>
    <description>The latest articles on DEV Community by Alex Hu (@alexhu0110).</description>
    <link>https://dev.to/alexhu0110</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%2F4246817%2F77adbd61-79c0-4449-b78d-c8f3616a0e32.png</url>
      <title>DEV Community: Alex Hu</title>
      <link>https://dev.to/alexhu0110</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/alexhu0110"/>
    <language>en</language>
    <item>
      <title>Henge: an open-source workspace for AI video character replacement</title>
      <dc:creator>Alex Hu</dc:creator>
      <pubDate>Mon, 16 Nov 2026 03:53:28 +0000</pubDate>
      <link>https://dev.to/alexhu0110/henge-an-open-source-workspace-for-ai-video-character-replacement-48a1</link>
      <guid>https://dev.to/alexhu0110/henge-an-open-source-workspace-for-ai-video-character-replacement-48a1</guid>
      <description>&lt;p&gt;A video character replacement has several inputs worth inspecting before the final generation: the source performance, the character reference, and the guidance that tells the model which subject to change.&lt;/p&gt;

&lt;p&gt;I built &lt;strong&gt;Henge&lt;/strong&gt;, an open-source workspace that puts those stages into a visible prepare → review → generate workflow.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://github.com/Hy-1990/Henge/blob/main/docs/assets/henge-x-en.mp4" rel="noopener noreferrer"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F2j9aplccslctjuupt3kz.gif" alt="Henge source and replacement comparison" width="540" height="400"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://github.com/Hy-1990/Henge" rel="noopener noreferrer"&gt;Source code and installation&lt;/a&gt; · &lt;a href="https://raw.githubusercontent.com/Hy-1990/Henge/main/docs/assets/henge-x-en.mp4" rel="noopener noreferrer"&gt;Watch the 20-second demo&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The demo compares a source clip with the actual result of one replacement run, using original AI-generated actors. The promotional edit omits processing time; the repository also includes the uncut result.&lt;/p&gt;

&lt;h2&gt;
  
  
  Prepare, inspect, then generate
&lt;/h2&gt;

&lt;p&gt;Start with a short source video and a character reference. Henge prepares motion guidance, lets you inspect it, and ties approval to that version of the inputs.&lt;/p&gt;

&lt;p&gt;There are three guidance methods: Depth + SAM, Fast Face Mesh, and a combination of Depth + SAM + Face Mesh. The face-mesh methods require an optional worker and are aimed at a single centered speaker.&lt;/p&gt;

&lt;p&gt;During review, check the full clip for hair, hands, clothing, mask coverage, and spill into the background. Changing the prepared input invalidates the earlier approval. Once the inputs are ready, submit the final generation, preview the output, and export with the source video's original audio restored locally.&lt;/p&gt;

&lt;p&gt;The application runs on your machine, while AI processing uses paid cloud services. You need your own &lt;strong&gt;DreamAPI and Replicate keys and available credit&lt;/strong&gt;. Guidance preparation can incur charges too; review is a checkpoint before final generation, not a promise of a free preparation stage.&lt;/p&gt;

&lt;h2&gt;
  
  
  Recovery is part of the workflow
&lt;/h2&gt;

&lt;p&gt;A long-running generation can finish even after the browser closes or the local connection breaks. Henge saves provider task IDs so reopening a task can reconnect to work that has already been submitted.&lt;/p&gt;

&lt;p&gt;An uncertain submission blocks automatic resubmission. A timeout without a task ID needs a check of provider history and billing before deliberately starting another paid job. This makes the state of a task visible instead of treating every interruption as a reason to generate again.&lt;/p&gt;

&lt;p&gt;Keys entered in the application use macOS Keychain or Windows Credential Manager. Source derivatives and references are processed by the selected cloud providers; this is a local interface, not an offline generation pipeline.&lt;/p&gt;

&lt;h2&gt;
  
  
  What is open source
&lt;/h2&gt;

&lt;p&gt;Henge is the workflow and interface. It does not include an open-source video generation model.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://api.newportai.com/" rel="noopener noreferrer"&gt;DreamAPI&lt;/a&gt; provides the Seedance 2.5 generation integration. Replicate supplies preparation services including depth, vocal separation, and the current SAM adapter. The repository documents those dependencies and their terms.&lt;/p&gt;

&lt;p&gt;Henge is derived from &lt;a href="https://github.com/sirioberati/Genjustsu-Open-Source-Workflow" rel="noopener noreferrer"&gt;Sirio Berati's Genjustsu Open Source Workflow&lt;/a&gt;, with the upstream MIT license retained. Upstream example videos and branding are not distributed with Henge.&lt;/p&gt;

&lt;h2&gt;
  
  
  Try a short, single-subject clip
&lt;/h2&gt;

&lt;p&gt;On macOS, you need Python 3.12+ and Rubber Band:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/Hy-1990/Henge.git
&lt;span class="nb"&gt;cd&lt;/span&gt; Henge
brew install python@3.12 rubberband
&lt;span class="nv"&gt;PYTHON&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;python3.12 bash setup.sh
bash start.command
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Open &lt;strong&gt;&lt;a href="http://127.0.0.1:8771" rel="noopener noreferrer"&gt;http://127.0.0.1:8771&lt;/a&gt;&lt;/strong&gt; and configure the two provider keys. Windows setup instructions are also included in the repository.&lt;/p&gt;

&lt;p&gt;Start with a clearly visible single subject in a 4–30 second clip. Automatic identity matching across cuts and pixel-perfect preservation of every non-target person are not guaranteed. A live macOS Depth + SAM / image-reference / 720p workflow has been verified; other modes and resolutions still need their own quality checks, and a paid Windows generation run has not yet been verified.&lt;/p&gt;

&lt;p&gt;The &lt;a href="https://github.com/Hy-1990/Henge" rel="noopener noreferrer"&gt;repository&lt;/a&gt; includes the source clip, character reference, uncut output, setup instructions, and current limitations.&lt;/p&gt;

</description>
      <category>showdev</category>
      <category>opensource</category>
      <category>ai</category>
      <category>python</category>
    </item>
    <item>
      <title>Flare: take a script to a short film, with control over every shot</title>
      <dc:creator>Alex Hu</dc:creator>
      <pubDate>Mon, 16 Nov 2026 03:51:58 +0000</pubDate>
      <link>https://dev.to/alexhu0110/flare-take-a-script-to-a-short-film-with-control-over-every-shot-1d3a</link>
      <guid>https://dev.to/alexhu0110/flare-take-a-script-to-a-short-film-with-control-over-every-shot-1d3a</guid>
      <description>&lt;p&gt;The first draft of a short film is only the beginning. You still need to change a shot, fix its timing, replace an image, and see whether the sequence works.&lt;/p&gt;

&lt;p&gt;I built &lt;strong&gt;Flare&lt;/strong&gt;, an open-source workspace that brings script planning, editable storyboards, shot editing, preview, and export into one local application.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://github.com/Hy-1990/Flare/blob/main/docs/assets/flare-launch.mp4" rel="noopener noreferrer"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F5j2sdg51tbau8p7gdt2c.gif" alt="Flare launch preview" width="720" height="405"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;a href="https://github.com/Hy-1990/Flare" rel="noopener noreferrer"&gt;Source code and installation&lt;/a&gt; · &lt;a href="https://raw.githubusercontent.com/Hy-1990/Flare/main/docs/assets/flare-launch.mp4" rel="noopener noreferrer"&gt;Watch the 28-second demo&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  From script to an editable sequence
&lt;/h2&gt;

&lt;p&gt;Start with a script and shape it into a storyboard. Each shot has its own text, duration, composition, colors, and artwork. You can reorder, add, remove, or revise individual shots, then preview the sequence again.&lt;/p&gt;

&lt;p&gt;Local script splitting works without an API key. With your own DeepSeek key, you can use AI storyboarding and request a natural-language revision to one shot. Local splitting is a straightforward starting point; it does not understand or expand a story by itself.&lt;/p&gt;

&lt;p&gt;For visuals, import your own images or generate them through &lt;a href="https://api.newportai.com/" rel="noopener noreferrer"&gt;DreamAPI&lt;/a&gt;. DreamAPI also provides optional narration generation. The external generation calls use your provider balance; the workspace and project editing run locally.&lt;/p&gt;

&lt;h2&gt;
  
  
  Preview and export use the same renderer
&lt;/h2&gt;

&lt;p&gt;Flare shares a deterministic renderer between the browser preview and exported frames. Chromium renders the frames, and FFmpeg handles the final video export.&lt;/p&gt;

&lt;p&gt;Each export uses a snapshot of the project, so edits made afterward do not change the version already in the queue. The application supports queued exports, cancellation, and persistent projects.&lt;/p&gt;

&lt;p&gt;Exports include H.264 MP4 video, SRT subtitles, and editable project JSON. You can choose landscape, portrait, or square layouts at 720p or 1080p and 24 fps. Media files remain separate from the project JSON when moving a project between machines.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the style library provides
&lt;/h2&gt;

&lt;p&gt;Flare includes &lt;strong&gt;99 style references&lt;/strong&gt; with demo links and production guides. Eight have basic procedural renderers; the other 91 need imported or generated images. There are also three original procedural styles.&lt;/p&gt;

&lt;p&gt;A reference is a direction to work toward, rather than a guarantee that choosing its name reproduces the upstream demo. Image-assisted styles animate still images with composition and camera movement; they do not generate new video clips.&lt;/p&gt;

&lt;p&gt;The upstream catalog comes from &lt;a href="https://github.com/lemomo-ai/lemo-opuscar" rel="noopener noreferrer"&gt;Lemo-Opuscar&lt;/a&gt;. Its guides, media, and third-party assets have separate licenses, described in the repository's notices.&lt;/p&gt;

&lt;h2&gt;
  
  
  Try it locally
&lt;/h2&gt;

&lt;p&gt;You need Node.js 22.13+ and FFmpeg 7+ on your PATH:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/Hy-1990/Flare.git
&lt;span class="nb"&gt;cd&lt;/span&gt; Flare
npm install
npm run setup:browser
npm run build
npm start
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Open &lt;strong&gt;&lt;a href="http://127.0.0.1:4317" rel="noopener noreferrer"&gt;http://127.0.0.1:4317&lt;/a&gt;&lt;/strong&gt;. You can start with local storyboards and procedural exports without configuring a model provider.&lt;/p&gt;

&lt;p&gt;The current version supports 1–16 shots and 3–120 seconds per project. It has been validated on macOS Apple Silicon; Linux and Windows still need verification. Live DeepSeek planning, single-shot revision, and one DreamAPI image workflow have been checked. Live narration generation has not yet been verified.&lt;/p&gt;

&lt;p&gt;The application code is MIT licensed. Installation details, verification commands, and current limitations are in the &lt;a href="https://github.com/Hy-1990/Flare" rel="noopener noreferrer"&gt;GitHub repository&lt;/a&gt;.&lt;/p&gt;

</description>
      <category>showdev</category>
      <category>opensource</category>
      <category>ai</category>
      <category>webdev</category>
    </item>
  </channel>
</rss>
