Turn raw bullet points, concept ideas, and reference images into production-ready, dramaturgically structured video screenplays — pressed into strict single-line, timecoded prompts for MiniMax H3, Maestro 2.1.6, Kling AI, and Runway Gen-3.
🔗 GitHub: https://github.com/wobushannes/Drehbuch--Referenzanker-Generator
⚠️ IMPORTANT — LM STUDIO MUST BE RUNNING
This tool relies on a local LM Studio instance as its private Vision & Text AI backend. You MUST have LM Studio installed, running, and its local server enabled on port 1234 (http://localhost:1234/v1/chat/completions). No cloud API keys are required — everything runs 100% locally and privately. If LM Studio is NOT running, image analysis and prompt generation will fail. Load a vision-capable model in LM Studio before sending tasks. Get it here: https://lmstudio.ai/
WHAT IT IS
A high-precision, production-grade application for filmmakers, real estate marketers, architects, and content creators. It turns rough ideas into dramaturgically structured screenplays, broken down into contiguous timecoded windows (e.g. 14.000 seconds per window), each formatted as exactly one single line — 0 line breaks — to guarantee compatibility with video models that corrupt state on multi-line inputs.
KEY FEATURES
📐 Strict Single-Line Window Pressing (0 Line Breaks) — Every window prompt is pressed into exactly 1 line, with character limits, line-count validation, and real-time syntax checking. Includes contiguous timecodes like TIMECODE 00:00.000 to 00:14.000.
🧠 Local LM Studio Integration — Connects directly to your local LM Studio server for fully private vision & text analysis of uploaded images (persons, companion pets, facades, floorplans, props).
🗺️ Multimodal Floorplan & Architectural Analysis — Reads blueprints and floorplans (PNG, JPG, WebP), extracts room axes, foyers, kitchen islands, and sightlines, and translates spatial directions into continuous camera moves (e.g. "Steadicam dolly-in along sightline...").
🏷️ Identity & Character Anchoring — Locks character features (hair, eyes, age, clothing, facial structure) and prop details using Maestro 2.1.6 binding tags (@Subject1_, @Building1_, @Object1_, @Logo1_), preventing identity cross-bleeding and character cloning across multi-window sequences.
👆 1-Click Reference Chips — Extracts all active project references into clickable insert chips, with built-in presets (Resilience Coaching in the Mountain Forest, Floorplan-Guided Kitchen Walkthrough, Yacht & Pier Voyage, Penthouse Smart Home).
🌐 Human-Readable German Prompt Breakdown — A 1-click "Verständliches Deutsch" box for every technical English prompt, detailing scene action, camera motion, dialogue, 100mm macro close-ups, soundscapes, and CTA claims.
🌍 Bilingual GUI (DE / EN) — Full internationalization with a persistent language switcher.
WORKFLOW
1. Upload References — Drag & drop images into the Reference Manager, then click "Send All Tasks to LM Studio" to extract vision anchors.
2. Configure Story & Target Audience — Pick your audience and use 1-click chips to insert references into the bullet-point editor.
3. Generate & Press Single-Line Windows — Click "In Single-Line Windows pressen", verify 0 line breaks, check the German breakdown, and copy prompts into MiniMax H3 or Maestro.
CREDITS
📸 Instagram: https://www.instagram.com/mo_ment_e/
🧙♂️ AI Wizards: https://ai-wizards.de/
👨💻 Johannes Wobus: https://johannes-wobus.de/
Sample: