Qwen Thinking Prompt Refiner

This workflow uses Qwen Thinking (Qwen3-4B-Thinking) to transform a short, simple user idea into a fully refined, visually grounded, model-ready prompt.

The pipeline is designed for creators who want high-quality, descriptive prompts without tag-stuffing or abstract language. It enforces a structured reasoning process where the core subject, action, and visual constraints are preserved, then expanded into a clear, detailed scene description optimized for image or video generation models.

The refined output focuses strictly on visible elements only—such as appearance, pose, environment, lighting, material textures, and spatial composition—while avoiding emotional interpretation, metaphor, or meaningless quality tags. The result is a single, fluent English paragraph that reads naturally and can be plugged directly into downstream text-to-image or text-to-video workflows (including cinematic pipelines like LTX Video).

This workflow is especially useful for:

• Prompt polishing and enhancement

• Cinematic or illustrative scene generation

• Maintaining prompt consistency and visual clarity

• Avoiding unstable “tag salad” prompting styles

The final refined prompt is displayed using a text preview node for easy copy-paste into other workflows.