Scenography student asks how to get consistent, art-directed AI concept images from ChatGPT
Cazabal · reddit · 2026-07-26
The poster, a scenography/set-design student, asks whether ChatGPT can reliably produce art-directed concept images rather than generic AI-looking outputs.
What they are trying to achieve
- Early-stage brainstorming and visual development for set design
- Monumental, ambiguous spaces with strong materiality, textiles, controlled lighting, cinematic architectural photography, and restrained palettes
- Consistency across multiple images without style drift
Where they are stuck
- Text prompts alone often produce overly polished, plastic-looking results
- Successive “fixes” change one part of the image while breaking another
- They want a repeatable workflow, not a list of vague adjectives
Questions they ask
- Whether ChatGPT can currently hit this level of art direction consistently
- How to structure prompts across space, materials, lighting, camera, lens, composition, and constraints
- Whether sketches, renders, collages, or reference images work better than pure text
- How to keep a coherent visual language across a series
- Whether targeted edits work better than regenerating whole images
- Which mistakes make images look obviously AI-generated
- What free courses or guides can teach a proper workflow
More from Multimodal
- DeepSeek says AGI is the only mainline, and coding agents are the priority right now — FinanceYF5 · 2026-07-26
- Nightly Stroll shares a cinematic AI night scene with a moonlit figure — worldofjoonas · 2026-07-26
- Reddit users test SCAIL-2 (Wan2GP) with 5-person animation at 720p 24fps — wikid24 · 2026-07-26
- A ComfyUI workflow request combines object removal, upscaling and depth maps for CNC carving — AshokManker · 2026-07-26
- InVideo says its Agent One tops an independent AI video agent benchmark — gen_ericai · 2026-07-26
- Beginners want a local ComfyUI workflow for cinematic scenography concepts on 8 GB VRAM — Cazabal · 2026-07-26