Traditional generative AI tools limit creators to a single, text-heavy input box. Long paragraphs of descriptive text often lead to unpredictable results, because models struggle to weigh many instructions at once. BLOCKLORE redefines the process with visual prompting: decomposing a prompt into modular, interconnected nodes on an infinite canvas.
By separating creative intent into individual, logical steps, visual prompting gives you absolute control over the generation pipeline. Here are the foundational techniques.
1. Modular prompting: breaking down the formula
Visual prompting relies on breaking a complex scene into distinct components and assigning each one to its own Text node. Instead of writing a single prompt such as "a medieval warrior standing in a misty forest, highly detailed, cinematic lighting, slow pan," the elements are separated: a subject node defining only the character ("a medieval warrior in iron plate armor"), an environment node defining the setting ("a dense, misty pine forest at dawn"), and an aesthetic node defining style and lighting ("cinematic volumetric lighting, dark fantasy aesthetic, cool color palette").
Connecting these individual Text nodes to a single Image node lets the model process each instruction with optimal clarity, producing output that matches the creative vision far more closely.
2. Parent-child chains for style consistency
Maintaining visual consistency across generations is one of the hardest problems in generative media, and parent-child node relationships solve it. When an Image node produces a successful character render, that node becomes the parent. To generate a different scene featuring the same character, link the parent's output to the input port of a downstream child Image node.
The connection passes the original visual parameters downstream as a direct style reference. Attach a new Text node to the child to specify a new action, such as "holding a sword," and the model preserves the character's facial features and armor design while executing the new prompt.
3. Sequential workflows: from draft to animated video
Visual prompting extends well beyond static images. Because the canvas also houses Video, Audio, and Script nodes, an entire production timeline can be mapped out visually in a left-to-right sequence: a Script node outlines the storyboard, camera moves, and shot list; it links to a Text node that generates specific prompt variables for each shot; the Text node connects to an Image node that renders the static scene layout; the resulting image is routed into a Video node where the frame is animated with precise motion controls; and an Audio node joins the final Video node to overlay synchronized sound design.
Connecting node outputs to downstream inputs builds a directed acyclic graph that propagates prompting instructions automatically. Because every intermediate generation stays visible on the canvas as a persistent node, variations can be branched, reused, and compared side by side without losing progress.
4. Eliminating randomness with the Director Stage
Traditional prompting leaves camera positioning entirely to the model. The Director Stage introduces spatial visual prompts that remove that randomness: pose a 3D scene, map out exact camera movements such as a dolly zoom or a slow cinematic pan, and feed those spatial parameters into the active prompt nodes as physical constraints. The model then renders according to those coordinates, delivering frame-accurate, repeatable compositions.
With visual prompting, creators move from passive guesswork to active, structural orchestration. The infinite canvas turns every text prompt, style reference, and animation parameter into a tangible, connectable building block, making the whole process more organized, repeatable, and powerful.
