Documentation Index
Fetch the complete documentation index at: https://runcrate.ai/docs/llms.txt
Use this file to discover all available pages before exploring further.
Generate images using Stability AI’s Stable Diffusion models through the Runcrate API. No GPU setup, no model hosting — one API call, one image.
Available models
| Model | Quality | Speed | Best for |
|---|
| SD 3.5 Large | High | Medium | Production images, detailed scenes |
| SD 3 Medium | Good | Fast | Prototyping, high-volume generation |
Basic image generation
from runcrate import Runcrate
client = Runcrate(api_key="rc_live_YOUR_API_KEY")
image = client.models.generate_image(
model="stabilityai/stable-diffusion-3-5-large",
prompt="A serene Japanese garden in autumn, koi pond with stone bridge, "
"golden maple leaves, soft morning mist, photorealistic",
aspect_ratio="16:9",
)
image.data[0].save("japanese-garden.png")
Using negative prompts
from runcrate import Runcrate
client = Runcrate(api_key="rc_live_YOUR_API_KEY")
image = client.models.generate_image(
model="stabilityai/stable-diffusion-3-5-large",
prompt="Professional headshot of a business executive, studio lighting, neutral background",
negative_prompt="cartoon, illustration, anime, blurry, low quality, deformed",
aspect_ratio="1:1",
)
image.data[0].save("headshot.png")
Controlling generation parameters
from runcrate import Runcrate
client = Runcrate(api_key="rc_live_YOUR_API_KEY")
image = client.models.generate_image(
model="stabilityai/stable-diffusion-3-5-large",
prompt="Detailed architectural blueprint of a modern house, clean lines, cross-section view",
num_inference_steps=40, # default ~28, higher = more detail
guidance=7.5, # how closely to follow the prompt
seed=42, # reproducible results
)
image.data[0].save("blueprint.png")
Batch generation
from runcrate import Runcrate
from concurrent.futures import ThreadPoolExecutor
client = Runcrate(api_key="rc_live_YOUR_API_KEY")
products = [
{"name": "candle", "prompt": "Luxury scented candle in matte black jar, warm lighting, product photography"},
{"name": "notebook", "prompt": "Premium leather notebook on wooden desk, fountain pen, natural light"},
{"name": "mug", "prompt": "Ceramic coffee mug with steam, dark moody background, product shot"},
]
def generate(p):
image = client.models.generate_image(
model="stabilityai/stable-diffusion-3-5-large",
prompt=p["prompt"], aspect_ratio="1:1",
)
image.data[0].save(f"{p['name']}.png")
print(f"Saved {p['name']}.png")
with ThreadPoolExecutor(max_workers=4) as pool:
pool.map(generate, products)
SD 3.5 Large vs. SD 3 Medium
| SD 3.5 Large | SD 3 Medium |
|---|
| Detail | Fine textures, realistic lighting | Sufficient for most uses |
| Speed | Medium | Fast |
| Best for | Final production assets | Prototyping, high-volume |
Tips
- Negative prompts significantly improve quality — exclude common artifacts.
- Seed parameter makes results reproducible. Save it for prompt iteration.
- SD 3.5 Large for customer-facing images. SD 3 Medium for internal/prototype.
- Aspect ratio options:
1:1, 16:9, 9:16, 4:3, 3:2.
Next steps
