Generative AI Google Brain

vision-use-cases-batch-virtual-try-on

3 min read

Virtual Try-On: Batch Generation Pipeline

Source notebook

Repo path: vision/use-cases/batch_virtual_try_on.ipynb · Open on GitHub · intermediate

Batch-generates virtual try-on images from person and apparel inputs using Google GenAI SDK.

Summary

It teaches how to use the Google Gen AI SDK for Python with the Virtual Try-On model to generate images of people wearing product apparel. The workflow installs the SDK, authenticates in Colab, initializes a GenAI enterprise client with a Google Cloud project and region, accepts local uploads or public URLs, creates every person-product pairing, calls recontext_image, previews results, and downloads a ZIP.

Key code patterns

Initialize GenAI client

PROJECT_ID = os.environ.get("GOOGLE_CLOUD_PROJECT")
LOCATION = os.environ.get("GOOGLE_CLOUD_REGION", "us-central1")
client = genai.Client(
    enterprise=True,
    project=PROJECT_ID,
    location=LOCATION,
)

Configures the Google GenAI client for a project and region before calling the model.

Normalize local and URL images

if local_images:
    person_image_obj = Image.from_file(location=person_img)
    product_image_obj = Image.from_file(location=product_img)
else:
    person_response = requests.get(person_img)
    person_response.raise_for_status()
    person_image_obj = Image(image_bytes=person_response.content)

Shows the two supported input paths: uploaded files and URL-fetched image bytes.

Batch pairwise generation

for person_img in person_images:
    for product_img in product_images:
        generated_image = client.models.recontext_image(
            model="virtual-try-on-001",
            source=RecontextImageSource(
                person_image=person_image_obj,
                product_images=[ProductImage(product_image=product_image_obj)],
            ),
        )

Creates one try-on result for each person and product image combination.

Download generated results

with zipfile.ZipFile(zip_filename, "w", zipfile.ZIP_DEFLATED) as zip_file:
    for i, result in enumerate(results):
        image_data = result.image_bytes
        zip_file.writestr(f"result_{i + 1}.png", image_data)
files.download(zip_filename)

Packages generated image bytes into a timestamped ZIP for local download.

Models & APIs used

  • Models: virtual-try-on-001
  • APIs / services: Vertex AI, Virtual Try-On API, Cloud Storage
  • SDKs / libraries: google-genai, google.colab, matplotlib, numpy, requests, Pillow

When to use this

Use this pattern when you need to batch-generate virtual try-on outputs across multiple person and apparel image combinations.

Gotchas & caveats

  • Colab requires auth.authenticate_user() before using Google Cloud credentials.
  • The notebook requires an existing Google Cloud project and enabling the Agent Platform API with apiid aiplatform.googleapis.com.
  • PROJECT_ID must be set directly or through GOOGLE_CLOUD_PROJECT.
  • LOCATION defaults to us-central1 from GOOGLE_CLOUD_REGION when no region is set.
  • Choose only one input option per batch job: local images or public URLs.
  • Public image inputs must be one URL per line, not comma-separated values or local file paths.
  • The notebook adds time.sleep(2) between jobs to help with API quota limits.
  • Supported clothing is limited in the notebook to tops, bottoms, and footwear.

Best practices

  • Reads project and region from environment variables when notebook parameters are not set.
  • Displays input images before generation and displays generated images afterward for visual inspection.
  • Uses response.raise_for_status() when fetching public image URLs.
  • Tracks current job and total jobs while processing the Cartesian product of inputs.
  • Catches per-job exceptions and skips failed jobs instead of stopping the whole batch.
  • Adds a short delay between API calls to reduce quota pressure.
  • Writes generated outputs into a compressed ZIP file for download.

Graph View

Backlinks