Stable Diffusion 3.5 Large: 6 Prompt Tests (1024px)
Stable Diffusion 3.5 Large is positioned as a stronger general-purpose text-to-image model with better prompt adherence and typography. This post runs six prompts that stress text, product lighting, composition control, portraits, 3D style, and surreal illustration.
Model link
Test setup
- Size: 1024×1024
- Steps: 28
- Guidance scale: 7.0
- Scheduler: FlowMatchEulerDiscreteScheduler
- Negative prompt: “bad quality, lowres, blurry, watermark”
Test 1: Typography on a sign

This prompt checks if the model can keep letters clean and aligned. The overall sign look often lands well, but exact letter-perfect typography still needs a design tool for production work.
Test 2: Product photo + label text

SD 3.5 Large does a nice job with glass, reflections, and soft studio lighting. Label placement looks plausible, but small text can drift or deform.
Test 3: Simple spatial control (three objects)

This is a fast way to see prompt adherence. If the colors or positions swap, add stronger anchors like “left edge” and “right edge” and reduce extra scene details.
Test 4: Portrait lighting control

The lighting direction callouts can work well when the prompt stays simple. For sharper realism, specify background bokeh and avoid conflicting style words.
Test 5: Isometric 3D render style

Isometric prompts test composition discipline. The best results keep clean edges, consistent lighting, and a readable top-down angle.
Test 6: Surreal illustration (composition and mood)

This type of prompt suits SD 3.5 Large: a clear subject, strong mood, and simple scene geometry. It tends to deliver a cohesive image without overcomplicating the layout.
Quick takeaways
| What was tested | What worked | What to watch |
|---|---|---|
| Typography | Convincing sign and label layouts | Exact letter-perfect text still hard |
| Lighting | Studio reflections and portrait lighting cues | Too many style words can confuse the result |
| Prompt adherence | Simple composition instructions often follow | Add stronger anchors when positions drift |