Titan Image Generator alternative: product photo edits on Sume

Titan Image Generator v2 is Amazon's older image model on Bedrock. Its model page (read 2026-10-06) lists text-to-image, inpainting, outpainting, image variation, image conditioning, color-guided content, background removal and subject-consistency fine-tuning. If you are moving product photo work to Sume, most of that list has a counterpart, and two items do not.

What you lose

Fine-tuning a model on your own product photos is a Bedrock feature. The Sume Image API has no training route, so subject consistency comes from reference images. Send the same two or three product shots with every request.

A reference-driven product call

The request below sends three views of one product and asks for a lifestyle scene. It prints the billed cost.

import os, requests

views = ["front", "side", "top"]
refs = [{"type": "image_url",
         "image_url": {"url": f"https://example.com/kettle-{v}.jpg"}} for v in views]
r = requests.post("https://api.sume.com/v1/images", timeout=90,
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    json={"model": "openai/gpt-image-2.5",
          "prompt": "The kettle from the references on a kitchen counter, morning light. "
                    "Keep its shape and logo exact.",
          "input_references": refs})
print(r.status_code, r.json().get("usage"))