
popopo-99/zy-cinematic-realism
Productivity workflowThis page was translated automatically. Technical details may differ from the original. View the original
Turn simple scene ideas into restrained, physically believable cinematic AIGC prompts.
Overview
One Scene Master. Different models. Native prompts. Dream Maker is not a new image generation model, nor is it a "universal movie script" package. It is a cinematic visual workflow designed for ChatGPT/Codex: first lock in characters, story moments, actions, scenes, camera positions, composition, lighting, props, time, weather, and constraints, then compile native prompts for the target model. Scene Master → Creative Grammar → Model Compiler → Result Repair The external brand is ; To maintain installation path compatibility, automatic triggers, and explicit invocation, the technical name is always zy-cinematic-realism, and the call name is always $zy-cinematic-realism. OpenAI's image adaptation baseline has been updated to , while the architecture of Scene Master and Model Compiler remains unchanged. According to OpenAI's migration guidelines, the first round of comparisons should try to keep prompts, reference images, scene facts, image intent, and limitations consistent; After identifying specific failed variables, local fixes should be made, without proactively recreating based on version numbers. The official release notes focus on detail, natural light, textures, reference subject fidelity, precise editing, and multi-round retention, with latency compared to Images 2.0. This aspect ratio description does not mean the image quality is scaled up. Continue to use natural language visual brief, CHANGE ONLY / PRESERVE EXACT with clear reference image responsibilities. Standard GPT Image, OpenAI image, and ChatGPT image generation requests default to 2.5-compatible Adapter; Explicit GPT Image 2 requests still support legacy compatibility.
Overview (machine translated)
Chinese | English
Dream Maker · DREAM DIRECTOR
**Turn a single idea into a visual plan that can be generated, modified, and continued. **
Starting from story ideas, script excerpts, or reference images, the scene is built, the visual world is developed, and then prompts and local fixes suitable for the target model are output.
v2.5.0 Official Version · Download the official version · Get started in 60 seconds · Scenario Practice · See Case Studies · Installation Tutorial
The author’s previous AIGC work expresses silence after a failed attempt using character distance, seat occlusion, and actual lighting sources. Method direction demonstrated; not regenerated as v2.4 test results. View original image
This is a visual creation skill for ChatGPT/Codex. You will receive a reproducible prompt, and if necessary, add short visuals for decisions, dream cards, or continuous image group settings. It is responsible for the scheme and prompt words; The actual image generation requires the image generation function of your environment, and does not automatically call the API or lock the underlying model due to installation.
Technical names remain compatible with the call name: zy-cinematic-realism / $zy-cinematic-realism.
Get Started in 60 Seconds
After installation, copy this line and start:
At the very least, just tell them: who, where, what they’re doing right now, what they least want. If you have fixed shots, lighting, frames, or models, write them directly.
Then modify it naturally, without remembering the pattern name:
If you only want to copy the prompt: add “give the prompt only.” If you want to design in depth: require expanding the Scene Master, director’s four-axis, or continuity settings. It will continue to use the model already determined in the dialogue; If unknown and affects compile, it will only be asked once, or you can say “give the model a neutral version first.”
**Not installed yet? ** Watch the full installation tutorial. There is no Skills entry point; you can copy from the beginner version that includes chat; The basic chat version does not include the full director library or model parameter adapter.
Start with the materials you have in hand
-
Story / Script: “First, create the scene art concept for the director to discuss.” "/ “Pick key shots of changes in character relationships.” → Art concept discusses candidate designs for space, materials, character modeling, props, and lighting; Narrative key shots depict a specific moment
-
Idea: “After two detectives fail their interrogation, take the night bus back to the police station and give me a single-frame movie prompt.” → Short image decision with target model prompt
-
Reference: “Figure 1 is a reference color, Figure 2 is a reference composition; The content is changed to an elderly man fixing watches in the early morning.” " → Reference Responsibilities, Core Rules, and New Scenario Prompts
-
Fix: “Characters and composition are correct, but only fixes the lighting that looks too much like an advertisement.” → Up to three main issues and local fixes instructions
-
Photo Sets: “Make eight photos of female knights after work, keeping the face, armor, garage, and lighting all consistent.” → Continuity settings, shot lists, and changes for each shot
-
Change: “Scene unchanged, switch to Seedream.” " / “Only change camera positions.” → Transcoding or single-variable variant
-
Continue: “Organize the project handover card, recording accepted changes and next steps.” " → Can copy the complete project status to the new dialogue
This page was translated automatically. Technical details may differ from the original.
Recommended Tools
Try a different keyword or remove a filter.
Install
npx skillfish add popopo-99/zy-cinematic-realism