<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
	<channel>
		<title>Image Generation on Andy Bulka</title>
		<link>https://abulka.github.io/tags/image-generation/</link>
		<description>Recent content in Image Generation on Andy Bulka</description>
		<generator>Hugo</generator>
		<language>en</language>
		
		
		
		
			<lastBuildDate>Sun, 16 Aug 2026 00:02:56 +1000</lastBuildDate>
		
			<atom:link href="https://abulka.github.io/tags/image-generation/index.xml" rel="self" type="application/rss+xml" />
			<item>
				<title>Cinema</title>
				<link>https://abulka.github.io/projects/websites/cinema/</link>
				<pubDate>Sat, 15 Aug 2026 00:00:00 +0000</pubDate>
				<guid>https://abulka.github.io/projects/websites/cinema/</guid>
				<description>&lt;h2 id=&#34;cinema--text-to-cinematic-shot-lists&#34;&gt;Cinema — Text to Cinematic Shot Lists&lt;/h2&gt;&#xA;&lt;p&gt;Cinema transforms a written story into an &lt;strong&gt;illustrated, cinematic shot list&lt;/strong&gt;. Paste a story,&#xA;and a local LLM turns it into a structured breakdown of shots — subject, action, emotional&#xA;beat, lighting, camera note, color palette, location — which you can then render as images&#xA;through your own local image-generation backend. It is built entirely around &lt;strong&gt;local&lt;/strong&gt;&#xA;AI: Ollama for the language model and a configurable render backend (ComfyUI etc.), so nothing&#xA;leaves your machine.&lt;/p&gt;</description>
			</item>
			<item>
				<title>Muse</title>
				<link>https://abulka.github.io/projects/websites/muse/</link>
				<pubDate>Sat, 15 Aug 2026 00:00:00 +0000</pubDate>
				<guid>https://abulka.github.io/projects/websites/muse/</guid>
				<description>&lt;h2 id=&#34;muse--dynamic-prompt-generation-for-ai-image-apps&#34;&gt;Muse — Dynamic Prompt Generation for AI Image Apps&lt;/h2&gt;&#xA;&lt;p&gt;Muse is a toolkit for &lt;strong&gt;systematically varying AI image-generation prompts&lt;/strong&gt;. The core idea is&#xA;a dynamic-prompt engine: instead of typing one prompt, you write a &lt;em&gt;template&lt;/em&gt; containing&#xA;&lt;code&gt;{category}&lt;/code&gt; tokens, and Muse fills each token with a random pick from a large curated pool —&#xA;so the same template can produce thousands of distinct results. It plugs into two worlds: the&#xA;&lt;strong&gt;Draw Things&lt;/strong&gt; app on Apple Silicon (as a script) and &lt;strong&gt;ComfyUI&lt;/strong&gt; workflows (local or cloud).&lt;/p&gt;</description>
			</item>
	</channel>
</rss>
