Titan Image Generator v2: 1,408 px and 5 MB limits vs Sume edits

Titan Image Generator v2 on Amazon Bedrock is strict about input size. If your product photos come from a DSLR or a phone, they are too big for it. This page gives the limits from Amazon's model page and a preflight check, then covers what Sume asks of an edit input.

Preflight before Bedrock

Resize first, then check the byte size. The function below returns the image bytes as a JPEG under the 5 MB cap or raises.

import io
from PIL import Image

MAX_SIDE, MAX_BYTES = 1408, 5 * 1024 * 1024

def titan_ready(path: str) -> bytes:
    im = Image.open(path).convert("RGB")
    im.thumbnail((MAX_SIDE, MAX_SIDE))
    for q in (95, 90, 85, 80):
        buf = io.BytesIO()
        im.save(buf, "JPEG", quality=q)
        if buf.tell() <= MAX_BYTES:
            return buf.getvalue()
    raise ValueError("still over 5 MB")

print(len(titan_ready("photo.jpg")))

What Sume wants

Sume takes edit inputs as public HTTPS URLs in input_references, not as base64 in the body. The number of references depends on the model: up to 10, 16 on GPT Image 2.5, and 5 on Ideogram 4.5 (the first is the image to edit). The Sume docs do not publish one pixel or byte cap for all models, so test one full-size photo per model before a batch run.

If the URL is private, the call fails with an input-media error. The public HTTPS checklist lists what to check.