<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>direktor — build notes</title><description>Pipeline-based AI video generation. Build notes from the team behind direktor (Skelf-Research).</description><link>https://direktor.skelfresearch.com/</link><language>en-gb</language><item><title>From script to shot list: where pipelines win over single-shot prompts</title><link>https://direktor.skelfresearch.com/blog/script-to-shot-list/</link><guid isPermaLink="true">https://direktor.skelfresearch.com/blog/script-to-shot-list/</guid><description>Why decomposing text-to-video into discrete stages outperforms throwing the whole task at a single model — and what direktor learned building it that way.</description><pubDate>Thu, 20 Nov 2025 00:00:00 GMT</pubDate></item><item><title>Why text-to-video is mostly text-to-storyboard right now</title><link>https://direktor.skelfresearch.com/blog/text-to-video-is-storyboard/</link><guid isPermaLink="true">https://direktor.skelfresearch.com/blog/text-to-video-is-storyboard/</guid><description>A frank look at what current AI video tools actually generate — stills with implied motion — and why naming that honestly makes for better tools and better expectations.</description><pubDate>Thu, 13 Nov 2025 00:00:00 GMT</pubDate></item><item><title>Compositing AI-generated shots without losing continuity</title><link>https://direktor.skelfresearch.com/blog/compositing-without-continuity-loss/</link><guid isPermaLink="true">https://direktor.skelfresearch.com/blog/compositing-without-continuity-loss/</guid><description>What breaks when you stitch independently generated images into something that needs to feel like one piece — and the practical tricks that actually help.</description><pubDate>Thu, 06 Nov 2025 00:00:00 GMT</pubDate></item></channel></rss>