<?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: theheaton hsu</title>
    <description>The latest articles on DEV Community by theheaton hsu (@theheaton_hsu_7420a2ac34c).</description>
    <link>https://dev.to/theheaton_hsu_7420a2ac34c</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%2F4248726%2F30ac68e1-a7f0-4104-a919-7f23bc6d846f.png</url>
      <title>DEV Community: theheaton hsu</title>
      <link>https://dev.to/theheaton_hsu_7420a2ac34c</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/theheaton_hsu_7420a2ac34c"/>
    <language>en</language>
    <item>
      <title>How I built an AI rap video generator as a solo maker</title>
      <dc:creator>theheaton hsu</dc:creator>
      <pubDate>Tue, 17 Nov 2026 02:45:15 +0000</pubDate>
      <link>https://dev.to/theheaton_hsu_7420a2ac34c/how-i-built-an-ai-rap-video-generator-as-a-solo-maker-1i52</link>
      <guid>https://dev.to/theheaton_hsu_7420a2ac34c/how-i-built-an-ai-rap-video-generator-as-a-solo-maker-1i52</guid>
      <description>&lt;p&gt;A few weeks ago my feed was full of the "Hotel Lobby" rap edits: two people in an orange booth, one hanging mic, rapping a famous song with swapped-in celebrity faces. I wanted a version where you use your own photo and get an original rap instead. So I built one. Here's how the pipeline works and what I learned.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it does
&lt;/h2&gt;

&lt;p&gt;You upload one photo (or two for a duo), pick a scene and a beat, and type a topic like "my golden retriever who steals socks". A few minutes later you get a 5–15 second MP4 where the face raps original lyrics in sync with a new beat.&lt;/p&gt;

&lt;h2&gt;
  
  
  The pipeline
&lt;/h2&gt;

&lt;p&gt;The job is split into small steps so each one can fail and retry on its own:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Moderation first.&lt;/strong&gt; The topic and any custom lyrics are checked before anything expensive runs. Public figures and people who haven't given permission are not allowed.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Lyrics.&lt;/strong&gt; An LLM (Anthropic's Claude) writes bars from the topic. The prompt fixes the number of lines to the video length, so a 5-second clip doesn't get 16 bars. Users can also paste their own lines, one per bar.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Beat, vocals and video.&lt;/strong&gt; The lyrics, the beat style (Trap, Boom Bap, Drill, Old School) and the scene go to ByteDance's Seedance 2.0 family, which generates the music, the voice and the lip-synced video in one pass.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Duo mode.&lt;/strong&gt; Two photos become two performers in one frame, trading lines. Getting the model to keep both faces and not invent extra people took more prompt work than anything else.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Delivery.&lt;/strong&gt; The finished file goes to object storage and the user downloads an MP4 in 9:16, 1:1 or 16:9.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Renders take 2–4 minutes, so the frontend polls a task record instead of holding a request open.&lt;/p&gt;

&lt;h2&gt;
  
  
  Credits and failures
&lt;/h2&gt;

&lt;p&gt;Video generation costs real money per second, so I use credits instead of a flat subscription. Credits are deducted when a task starts and refunded automatically if the render fails, so nobody has to email me to get credits back for a broken video.&lt;/p&gt;

&lt;h2&gt;
  
  
  Stack
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;TanStack Start (React 19, Vite) on Cloudflare Workers&lt;/li&gt;
&lt;li&gt;Drizzle ORM, better-auth&lt;/li&gt;
&lt;li&gt;A small task table that tracks each AI job, its cost and its status&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  What I'd do differently
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Ship a free tool early.&lt;/strong&gt; A free AI rap lyrics generator (no account needed) is an easy way for people to try the idea before they pay for a video.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Test with pets early.&lt;/strong&gt; People love making their dog rap. Faces that aren't human break assumptions in prompts, so test them from day one.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Keep scenes few and good.&lt;/strong&gt; Seven scenes with tuned prompts beat thirty generic ones.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;If you want to try it, it's at &lt;a href="https://rapvideo-ai.io" rel="noopener noreferrer"&gt;rapvideo-ai.io&lt;/a&gt;. New accounts get free credits for a first short video. I'd love feedback on the pipeline, especially from anyone who has fought with lip-sync on non-human faces.&lt;/p&gt;

</description>
      <category>webdev</category>
      <category>showdev</category>
      <category>buildinpublic</category>
    </item>
  </channel>
</rss>
