---
title: Prompting for photorealism, illustration, and design — Recraft V4.1 | Runware Docs
url: https://runware.ai/docs/models/recraft-v4-1/guides/prompting
description: How to write effective prompts for Recraft V4.1, from short interpretive prompts to structured multi-layer descriptions for precise creative control.
---
### [Introduction](https://runware.ai/docs/models/recraft-v4-1/guides/prompting#introduction)

Most image generation models need long, specific prompts to produce anything usable. Short inputs tend to produce generic, flat results. V4.1 works differently: it reads **short prompts** better than previous versions, filling in composition, lighting, and stylistic choices that produce **visually rich, emotionally engaging** output by default. It also follows long **structured prompts** with precision when you need exact control.

This means you can work in two modes. Use short prompts when you're exploring ideas and want the model to make aesthetic decisions. Use structured prompts when you have a specific composition in mind and need every element placed intentionally.

![A cinematic portrait of a woman in golden hour light, wind moving through her hair, shot with shallow depth of field](https://runware.ai/docs/assets/hero.KdgFAPS5_ZLU6Qj.jpg)

> **Prompt**: golden hour portrait

The image above was generated from a two-word prompt. V4.1 chose the framing, lighting direction, depth of field, and color grade on its own.

This guide covers both approaches, then walks through the visual formats where V4.1 performs strongest: photorealism, illustration, 3D rendering, and graphic design.

> [!NOTE]
> The prompting techniques in this guide apply to the entire Recraft model family: [Recraft V4.1 Pro](https://runware.ai/docs/models/recraft-v4-1-pro), [Recraft V4.1 Utility](https://runware.ai/docs/models/recraft-v4-1-utility), [Recraft V4.1 Utility Pro](https://runware.ai/docs/models/recraft-v4-1-utility-pro), and the [V4 generation](https://runware.ai/docs/models/recraft-v4) including its [Vector](https://runware.ai/docs/models/recraft-v4-vector) and [Pro](https://runware.ai/docs/models/recraft-v4-pro) variants.
> 
> **When to use Utility over the main model:** Recraft V4.1 brings its own creative point of view to every prompt - it experiments with angle, light, and composition. The Utility variants do the opposite: flat lighting, front-facing composition, and simple scenes. Use Utility for mockups, product shots, icon sets, or design system assets where predictable output matters more than creative surprise.

### [Short prompts and interpretive mode](https://runware.ai/docs/models/recraft-v4-1/guides/prompting#short-prompts-and-interpretive-mode)

When you give V4.1 a prompt under **~15 words**, the model enters what Recraft calls **interpretive mode**. Instead of following a rigid instruction, it makes its own compositional decisions: where to place the subject and how to light the scene.

![An autumn forest path with fallen leaves and fog between the trees](https://runware.ai/docs/assets/short-1.D-i_4nDO_KLeog.jpg)

*autumn forest path*

> **Prompt**: autumn forest path

![A cup of coffee on a rainy windowsill with water droplets on the glass](https://runware.ai/docs/assets/short-2.Crtku9fx_Zkhzvz.jpg)

*coffee on a rainy day*

> **Prompt**: coffee on a rainy day

Both images used prompts under five words. The model decided on composition, color temperature, camera angle, and atmospheric effects. V4.1 defaults to high-aesthetic output: it picks lighting and mood that feel emotionally intentional rather than neutral.

Short prompts are effective for:

- **Early-stage exploration** when you're looking for a direction, not a specific result
- **Mood discovery**, where you want to see how the model interprets an abstract concept
- **Concept sketching** before investing time in detailed prompt engineering
- **Generating visual references** for a brief or pitch deck

TypeScriptPythoncURLCLIJSON

```typescript
import { createClient } from '@runware/sdk'

const client = await createClient({ apiKey: process.env.RUNWARE_API_KEY })
await client.connect()

const [result] = await client.run({
  model: 'recraft:v4.1@0',
  positivePrompt: 'coffee on a rainy day',
  width: 1024,
  height: 1024
})
```

```python
import asyncio
import os

from runware import Runware

async def main():
    async with Runware(api_key=os.environ["RUNWARE_API_KEY"]) as client:
        results = await client.run({
            "model": "recraft:v4.1@0",
            "positivePrompt": "coffee on a rainy day",
            "width": 1024,
            "height": 1024
        })

asyncio.run(main())
```

```bash
curl https://api.runware.ai/v1 \
  -H "Authorization: Bearer $RUNWARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '[
    {
      "taskType": "imageInference",
      "taskUUID": "d4e5f6a7-b8c9-0123-def0-456789012345",
      "model": "recraft:v4.1@0",
      "positivePrompt": "coffee on a rainy day",
      "width": 1024,
      "height": 1024
    }
  ]'
```

```bash
runware run recraft:v4.1@0 \
  positivePrompt="coffee on a rainy day" \
  width=1024 \
  height=1024
```

```json
{
  "taskType": "imageInference",
  "taskUUID": "d4e5f6a7-b8c9-0123-def0-456789012345",
  "model": "recraft:v4.1@0",
  "positivePrompt": "coffee on a rainy day",
  "width": 1024,
  "height": 1024
}
```

### [Structured prompts for precise control](https://runware.ai/docs/models/recraft-v4-1/guides/prompting#structured-prompts-for-precise-control)

When you need specific results, V4.1 follows detailed prompts with **high fidelity**. The key is organizing your prompt in layers, from the most important element to the least:

1. **Subject and scene** - who or what is in the image
2. **Environment** - where the scene takes place
3. **Framing and pose** - how the subject is positioned relative to the camera
4. **Physical details** - identity, clothing, materials, textures
5. **Lighting** - direction, quality, color temperature
6. **Camera** - angle, depth of field, lens characteristics
7. **Mood** - overall emotional tone and atmosphere

**[Subject]** A traditional Japanese zen garden with a red wooden torii gate at the end of a curved stone path through raked white gravel, **[Environment]** mature cherry blossom trees on both sides with petals falling into a koi pond on the left, **[Framing]** centered composition from the path's entrance, **[Details]** moss on the gate posts and morning mist settling low, **[Lighting]** soft overcast light, **[Camera]** shot at eye level with a 50mm lens, **[Mood]** serene and meditative atmosphere

Compare the structured prompt above with a short one for the same concept. The short prompt gives you the model's interpretation. The structured prompt gives you yours.

![A Japanese garden in spring with cherry blossoms](https://runware.ai/docs/assets/minimal.Dxv_YT-8_ZFe6Cw.jpg)

*Short: "Japanese garden in spring"*

> **Prompt**: Japanese garden in spring

![A traditional Japanese zen garden with a torii gate, cherry blossoms, and morning mist in a centered composition](https://runware.ai/docs/assets/detailed.CoBOG3K0_Z2tyhgr.jpg)

*Structured: full scene description*

> **Prompt**: A traditional Japanese zen garden with a red wooden torii gate at the end of a curved stone path through raked white gravel, mature cherry blossom trees on both sides with petals falling into a koi pond on the left, centered composition from the path's entrance, moss on the gate posts and morning mist settling low, soft overcast light, shot at eye level with a 50mm lens, serene and meditative atmosphere

You don't need all seven layers in every prompt. Use as many as the image requires. A product shot might only need subject, lighting, and camera. A character illustration might need all seven.

![A street food vendor in Bangkok at night serving pad thai from a lit cart, warm tungsten lighting mixing with neon signs](https://runware.ai/docs/assets/structured-1.CaZ8fA95_1Ga6Jn.jpg)

*Structured: scene + environment + lighting*

> **Prompt**: A street food vendor serving pad thai from a wheeled cart at night in Bangkok, crowded narrow alley with other food stalls visible behind, warm tungsten bulb hanging from the cart's canopy mixing with cool blue neon signs in the background, steam rising from the wok, shot from slightly below at 35mm, shallow depth of field on the vendor's hands, bustling atmosphere

![An architect at a drafting table examining blueprints under a single desk lamp in a mid-century modern office](https://runware.ai/docs/assets/structured-2.BQQM08Ao_Z1DBGAp.jpg)

*Structured: subject + environment + mood*

> **Prompt**: A female architect in her 40s examining blueprints at a large wooden drafting table, mid-century modern office with floor-to-ceiling bookshelves and a large arched window showing a city skyline at dusk, single brass desk lamp casting warm directional light, reading glasses pushed up on her forehead, concentrated expression, muted earth tones, contemplative and focused atmosphere

### [Visual strengths by format](https://runware.ai/docs/models/recraft-v4-1/guides/prompting#visual-strengths-by-format)

V4.1 handles multiple visual formats **without switching models**. The same API call produces different styles based on your prompt language.

#### [Photorealism](https://runware.ai/docs/models/recraft-v4-1/guides/prompting#photorealism)

V4.1's photorealistic output looks photographed rather than generated. Backgrounds are quieter and lighting is more purposeful. Skin and material textures read as natural.

Describe the scene as if you're briefing a photographer: specify the lens, the lighting setup, and the environment.

![A woman reading a book at a sunlit cafe table with an espresso and croissant](https://runware.ai/docs/assets/photo-lifestyle.B_XnVP19_tiwxO.jpg)

*Lifestyle photography*

> **Prompt**: A young woman in a cream linen blouse reading a paperback book at a small marble cafe table, espresso cup and half-eaten croissant beside her, Parisian sidewalk cafe with green bistro chairs, morning sunlight casting long shadows across the table, shot at 85mm f/1.8 with natural bokeh from passing pedestrians, warm and relaxed editorial style

![A pair of wireless earbuds in a matte black charging case on a concrete surface](https://runware.ai/docs/assets/photo-product.DDkqxfk-_mzh25.jpg)

*Product photography*

> **Prompt**: A pair of wireless earbuds in a matte black oval charging case, sitting on a raw concrete surface, single soft studio light from the upper left creating a clean shadow to the right, product hero shot, top-down 45-degree angle, minimal background, premium tech aesthetic

#### [Illustration and 3D](https://runware.ai/docs/models/recraft-v4-1/guides/prompting#illustration-and-3d)

Illustration prompts benefit from specifying the drawing style, line behavior, and color logic explicitly. For 3D renders, describe the material and lighting as you would in a rendering engine.

![A hand-drawn illustration of a fox sitting in a forest clearing with watercolor textures](https://runware.ai/docs/assets/illustration.DYc-m4iE_Z1NsQBu.jpg)

*Illustration*

> **Prompt**: A hand-drawn illustration of a red fox sitting upright in a mossy forest clearing, loose watercolor washes with visible paper texture, ink outlines with varying stroke weight, muted greens and warm oranges, children's book illustration style, whimsical but not cartoonish, small mushrooms and ferns in the foreground

![A 3D designer toy bear character in matte vinyl with pastel colors on a pedestal](https://runware.ai/docs/assets/3d-toy.Ck3UG53K_Z1hShlH.jpg)

*3D render*

> **Prompt**: A 3D designer toy of a small bear character in matte vinyl finish, pastel pink body with mint green overalls, standing on a white circular pedestal, smooth rounded proportions with minimal detail, soft studio lighting from three directions, light gray seamless background, collectible figurine product shot

#### [Graphic design and logos](https://runware.ai/docs/models/recraft-v4-1/guides/prompting#graphic-design-and-logos)

For posters and graphic layouts, describe the typographic hierarchy and compositional mechanics. For logos, specify the shape logic, color system, and constraints (no gradients, no texture, etc.).

![A retro-style travel poster for Tokyo with bold typography and a stylized cityscape](https://runware.ai/docs/assets/poster.bzQMKgD3_ZLK4N.jpg)

*Poster design*

> **Prompt**: A retro travel poster for Tokyo, bold sans-serif text reading "TOKYO" at the top in white, stylized illustration of Tokyo Tower and Mount Fuji in the background, limited color palette of coral red, navy blue, and cream, flat graphic style with halftone dot texture, vintage print aesthetic, vertical format

![A set of four minimalist animal logos in a consistent geometric style](https://runware.ai/docs/assets/logo-set.DsCH3w9D_1EIsFy.jpg)

*Logo system*

> **Prompt**: A set of four minimalist animal logos arranged in a 2x2 grid: an owl, a whale, a deer, and a fox, each constructed from simple geometric shapes with consistent line weight, single color per icon using navy blue on white background, no gradients or shadows, professional brand identity style, scalable vector aesthetic

> [!NOTE]
> For vector and logo work, consider using [Recraft V4 Vector](https://runware.ai/docs/models/recraft-v4-vector) or [Recraft V4 Pro Vector](https://runware.ai/docs/models/recraft-v4-pro-vector), which are optimized for illustrated output with cleaner lines and flatter color fills.

### [Tips](https://runware.ai/docs/models/recraft-v4-1/guides/prompting#tips)

1. **Short prompts work best for photorealism and atmospheric scenes.** The model's default aesthetic choices lean toward natural lighting and balanced composition, which suits photographic styles well.
    
2. **Specify the medium early.** "A watercolor illustration of..." or "A 3D render of..." at the start of the prompt sets the visual format before the model starts interpreting the scene.
    
3. **Put text in quotation marks.** If your prompt includes text that should appear in the image (poster headlines, brand names, labels), always quote it. Without quotes, the model treats the words as scene description rather than literal content.
    
4. **Describe lighting like a photographer.** "Soft overcast light" produces very different results from "harsh midday sun" or "single tungsten bulb from above". V4.1 responds well to specific lighting direction and quality.
    
5. **Use lens references for photorealism.** Adding "shot at 85mm f/1.8" or "wide-angle 24mm" changes the spatial relationships and depth of field in ways the model handles accurately.