imagepipeline

On-model imagery for fast brands.

ImagePipeline is the image AI API for fashion and ecommerce. Turn flat-lays into on-model shots, run virtual try-on, and relight backgrounds — from one API. No studio, no photoshoot.

imagepipeline.io · Book a demo · Docs

Installation

pip install imagepipeline

Requires Python 3.9+ and a free API key from imagepipeline.io.

Building an AI agent?

There's a standalone MCP server — published separately as @imagepipeline/mcp (Node) — that exposes ImagePipeline as agent tools for Claude Desktop, Claude Code, or any MCP client. It's a separate tool, independent of this Python package; see its README to set it up.

Quick start

from imagepipeline import ImagePipeline

ip = ImagePipeline("ip_live_xxxxxxxxxxxx")

# Turn a flat-lay into an on-model shot
result = ip.generate.image(
    prompt="fashion model wearing the product, white studio background, editorial lighting"
)
print(result.url)

Core use cases

Virtual try-on

Dress a real person in any clothing item. Pass the model photo and the product image — ImagePipeline handles the compositing.

result = ip.identity.tryon(
    person_image="https://cdn.example.com/model.jpg",
    clothing_image="https://cdn.example.com/shirt.jpg",
    gender="woman",
)
print(result.url)

Background replacement

Swap a flat studio background for any scene. The subject is automatically isolated — no masking required. Describe the subject so it's preserved cleanly.

result = ip.background.change(
    input_image="https://cdn.example.com/product-shot.jpg",
    prompt="minimal white marble surface, soft natural light",
    subject_description="glass perfume bottle",
)
print(result.url)

Background removal

Cut out the subject as a transparent PNG, or recolor onto a flat background with an optional drop shadow.

# Transparent cutout
job = ip.background.remove(input_image="https://cdn.example.com/product.jpg")
print(job.cutout_url)

# Flat background + drop shadow
job = ip.background.remove(
    input_image="https://cdn.example.com/product.jpg",
    recolor="#FFFFFF",
    drop_shadow=True,
)
print(job.url)

On-model image generation

Generate consistent AI models with your brand's look and feel using identity profiles.

# Create a reusable brand model profile
profile = ip.identity.create_profile(
    name="Campaign Model",
    prompt_template="{{ user_prompt }}, Caucasian woman, late 20s, blue eyes, studio lighting",
    seed_strategy="fixed",
    fixed_seed=42,
)

# Generate on-model imagery with consistent identity
result = ip.generate.image(
    prompt="wearing a navy linen blazer, white background",
    profile_id=profile["profile_id"],
)
print(result.url)

Image editing

Edit product images with natural-language instructions.

result = ip.edit.image(
    input_image="https://cdn.example.com/product.jpg",
    prompt="remove the wrinkles from the fabric, keep everything else identical",
)
print(result.url)

Upload & edit

Upload local assets and use them directly in any editing workflow.

# Upload a flat-lay image
upload = ip.upload.image("flat-lay.jpg")

# Then use it as an edit input
result = ip.edit.image(
    input_image=[upload.url, "https://cdn.example.com/model.jpg"],
    prompt="put the model in the flat-lay clothing",
    mask_segment="upper-clothes",
)
print(result.url)

Targeted editing with segmentation

Use segment detection to edit only the clothing — face, hair, and background stay pixel-perfect.

# Optional: preview what segments exist in the image
seg = ip.segment.image("https://cdn.example.com/model.jpg")
for s in seg.segments:
    print(s.label, "→", s.display)
# upper-clothes → Top / Shirt
# pants         → Pants

# Edit only the detected segment
result = ip.edit.image(
    input_image=["https://cdn.example.com/model.jpg", "https://cdn.example.com/new-shirt.jpg"],
    prompt="dress the model in the shirt from image 2",
    mask_segment="upper-clothes",   # face, background, pants unchanged
)
print(result.url)

Async & webhooks

All methods default to wait=True (blocks until the job completes). For high-throughput pipelines:

# Fire-and-forget — returns immediately
job = ip.generate.image(prompt="...", wait=False)
print(job.job_id)

# Poll manually when ready
completed = ip._transport.poll("generate/image/v1", job.job_id)
print(completed.url)

# Or use a webhook — we POST a WebhookEvent to your URL on completion
job = ip.identity.tryon(
    person_image="https://...",
    clothing_image="https://...",
    callback_url="https://yourserver.com/hooks/imagepipeline",
    wait=False,
)

Error handling

from imagepipeline.exceptions import (
    AuthenticationError,
    RateLimitError,
    JobFailedError,
    JobTimeoutError,
    APIError,
)

try:
    result = ip.generate.image(prompt="...")
except AuthenticationError:
    print("Invalid API key — get one at imagepipeline.io")
except RateLimitError as e:
    print(f"Rate limited — retry after {e.retry_after}s")
except JobFailedError as e:
    print(f"Job {e.job_id} failed: {e.reason}")
except JobTimeoutError as e:
    print(f"Timed out after {e.timeout}s")
except APIError as e:
    # e.error_code is the stable machine-readable code, e.g. "INSUFFICIENT_BALANCE",
    # "INVALID_PARAMETERS", "NOT_FOUND" — use it for programmatic handling.
    print(f"HTTP {e.status_code} ({e.error_code}): {e}")

API reference

Requirements

• Python 3.9+
• requests

License

MIT — see LICENSE.

imagepipeline.io · Book a demo · Docs