Z-Image Turbo Prompting Guide

What Z-Image Turbo is (and isn't)
Z-Image Turbo is a 6B single-stream diffusion transformer (S3-DiT) that co-processes text and image tokens together. It runs in ~8 diffusion steps, optimised for fast, instruction-following generation with strong bilingual (English + Chinese) capability.
Critical difference from classic Stable Diffusion: The Turbo model runs at `guidance_scale = 0.0` and ignores the negative prompt field entirely — it doesn't use classifier-free guidance at inference. Any "negative prompt" box in a UI is cosmetic for this model.
Implication: every constraint lives in the positive prompt. Think of it as a very obedient camera crew: if you don't say it, it's allowed; if you say it vaguely, it will improvise.
---
Prompt scaffold
The recommended structure — all in the positive prompt:
``` [Shot type & angle] + [Subject & appearance] + [Clothing] + [Background/environment] + [Lighting] + [Mood] + [Style/medium] + [Technical quality] + [Safety/cleanup constraints] ```
Sweet spot: 80–250 words. Long and precise is good; long and novelistic/poetic is worse.
Sections to think through
| Section | Examples |
|---|---|
| Shot type | `close-up`, `medium shot`, `full-body`, `wide establishing shot` |
| Angle | `front view`, `45° angle`, `profile`, `looking slightly up` |
| Clothing | `dark blazer over a light shirt, fully clothed, modest professional outfit` |
| Background | `plain studio background`, `soft blurred vineyard at dusk`, `minimal interior` |
| Lighting | `soft diffused daylight`, `cinematic warm key light`, `golden-hour backlighting`, `studio portrait lighting` |
| Mood | `calm and professional`, `warmly inviting`, `dramatic`, `serene and pastoral` |
| Style | `realistic photograph`, `flat vector illustration`, `watercolor painting`, `oil painting` |
| Cleanup | `no text, no watermark, no logos, no extra limbs, correct human anatomy` |
---
Replacing negative prompts
Since the negative prompt field is ignored, embed constraints directly:
Artifacts:
- `correct human anatomy, natural hands and fingers, no extra limbs`
- `sharp focus, no motion blur, no grainy noise`
- `no text, no watermark, no logos, no UI elements`
- `simple uncluttered background, nothing distracting behind the subject`
Safety/modesty (when depicting people):
- Always include `adult` next to any human subject
- Describe clothing explicitly: `wearing a long-sleeved shirt and trousers, fully clothed`
- End with a safety clause: `safe for work, non-sexual, no nudity, no revealing clothing, no suggestive poses`
Stereotypes:
- Replace loaded tokens (`businessman`, `model`) with role + traits: `adult woman, office team lead, short dark hair, wearing casual jeans and a hoodie`
- Add: `realistic body types, no exaggerated proportions, respectful depiction`
---
Parameters
| Parameter | Recommended |
|---|---|
| `guidance_scale` | `0.0` (Turbo default — higher values rarely help) |
| `negative_prompt` | Leave blank (ignored by the model) |
| Steps | 8–12 |
| Resolution | 1024×1024 native; 768 fine for drafts |
| Seed | Fix while iterating a prompt; randomise when exploring variety |
| Max tokens | 512 (raise to 1024 for very long prompts) |
---
Drop-in templates (generic, safe)
Photorealistic professional headshot
> A close-up headshot of an adult woman in her 30s, friendly confident expression, medium-length dark hair, wearing a simple dark blazer over a light shirt, studio portrait setting, soft diffused daylight from the front, subtle blurred gray background, realistic photography, 85mm lens, shallow depth of field, detailed but natural skin, fully clothed, modest professional outfit, no logos, no text, no watermark, safe for work.
Group shot with diversity
> A medium-wide photo of five adult coworkers of diverse ethnicities and genders standing together in an open-plan office, casual smart outfits, soft natural daylight from large windows, relaxed but professional mood, realistic photography, realistic body types, respectful and non-sexual depiction, fully clothed, no revealing clothing, no stereotypes, no logos, no text, clean unobtrusive background, safe for work.
Flat vector illustration
> A flat vector illustration of an adult woman riding a bicycle through a park, stylized trees and simple buildings in the background, limited pastel color palette, clean modern design, no outlines, minimal shading, modest casual clothing, long-sleeved shirt and trousers, fully clothed, cheerful mood, clear blue sky, no text, no watermark, no logos, simple uncluttered background.
---
France project templates
Ready-to-adapt prompts tuned to this wiki's page types. Load API keys first: `set -a; source .secrets.env; set +a`.
Champagne / vineyard landscape
> A wide establishing shot of rolling Champagne vineyards in early autumn, rows of Chardonnay vines with pale yellow-green leaves on gently sloping chalk hills, a cluster of stone village rooftops and a church steeple visible in the middle distance, soft morning mist in the valley, warm golden-hour backlighting from the east, serene pastoral mood, realistic landscape photography, wide-angle lens, sharp foreground vines with hazy distance, no text, no watermark, no logos, no people.
Alsace village street
> A medium establishing shot of a half-timbered Alsatian village street in autumn, colorful flower boxes on carved wooden balconies, warm ochre and terracotta facades, uneven cobblestone lane receding into the distance, soft overcast daylight, cozy and inviting mood, realistic travel photography, 35mm lens, gentle lens compression, no people, no text, no watermark, no logos, no cars.
Wine cellar interior
> A wide interior shot of a French chalk wine cellar, rows of aging Champagne bottles stacked in riddling racks on both sides of a long stone-vaulted corridor, warm amber candle-and-spotlight lighting, subtle condensation on the walls, cool underground atmosphere, realistic photography, symmetrical composition, 24mm wide lens, deep focus, no people, no text, no watermark, no logos.
Winery / cellar door exterior
> A medium exterior shot of a French winery building, pale limestone facade with a wooden sign, climbing vines on the wall, open wooden doors hinting at a tasting room inside, terracotta roof tiles, soft afternoon sun casting long shadows, warm inviting mood, realistic travel photography, slight wide angle, no people, no text, no watermark, no logos.
Heritage village portrait (Mary's Alsace / Jerry's Moselle lines)
> A full-body photo of an adult woman in her 60s, silver-grey hair, modest warm-toned blouse and light jacket, standing on a cobblestone village square in front of a half-timbered Alsatian church, gentle autumn sunlight, candid travel style, soft background blur, realistic photography, 50mm lens, calm and reflective mood, fully clothed, no logos, no text, no watermark, safe for work.
Tasting scene (two people)
> A medium-wide photo of an adult man and an adult woman, both in their 60s, seated at a wooden tasting table in a sunlit French winery, each holding a stemmed white-wine glass to the light, a tasting notebook and a few open bottles on the table, warm golden window light, relaxed and curious mood, realistic candid photography, shallow depth of field, fully clothed casual-smart outfits, no logos, no watermark, no text, safe for work.
Itinerary mood shot (train travel)
> A wide-angle interior shot of a French TGV or regional train carriage, two seats by a large window, a glass of wine and a folded paper map on the fold-out tray, vineyards or rolling French countryside visible through the window at speed, warm afternoon light streaming in, calm and wandering mood, realistic travel photography, 28mm lens, no people, no text, no watermark, no logos.
---
Quick checklist before generating
1. Did I say "adult" next to every human subject? 2. Is clothing described explicitly (garment type + "fully clothed")? 3. Have I added a "safe for work / no nudity / non-sexual" clause if people appear? 4. Did I specify background simplicity and `no logos / no watermark / no extra text`? 5. Did I define angle, shot type, lighting, and mood clearly? 6. Is the negative prompt box in my UI actually doing anything for Turbo? (Almost certainly: no.)
---