Ad image ratios for LinkedIn, Pinterest, X and the Sume ratio to ask
Pinterest wants 2:3, X wants 1:1 or 1.91:1, LinkedIn's form image is 552x200. See which Sume aspect_ratio to request and what to crop.

Ask Sume for 2:3 for a Pinterest image pin, 1:1 or a cropped 2:1 for an X image card, and 3:1 for LinkedIn's Lead Gen Form image. The targets come from Pinterest's ad specs, X's Creatives documentation and LinkedIn's Lead Gen Form specifications; the ratio values Sume accepts are listed per model in its Image API.
Side by side
| Placement | Target | Sume `aspect_ratio` to request |
|---|---|---|
| Pinterest image pin | 2:3 (1000 x 1500 pixels) | 2:3 |
| Pinterest carousel | 1:1 or 2:3 | 1:1 or 2:3 |
| X website image card | 1:1 or 1.91:1, min 800px wide | 1:1, or 2:1 then crop |
| LinkedIn Lead Gen Form image | 552px by 200px (2.76:1) | 3:1 then crop |
Crop math in one function
When the ratio you need is not offered, request the nearest wider one and center-crop. This returns the box to crop to.
def center_crop_box(w: int, h: int, target: float) -> tuple[int, int, int, int]:
if w / h > target:
new_w = round(h * target)
left = (w - new_w) // 2
return (left, 0, left + new_w, h)
new_h = round(w / target)
top = (h - new_h) // 2
return (0, top, w, top + new_h)
print(center_crop_box(3000, 1000, 552 / 200))
print(center_crop_box(2000, 1000, 1.91))Check the model's list first
Ratio lists differ by model, so call GET /v1/images/models and read aspect_ratio for the one you use.
Sources
Related posts
More in Comparisons
- Ad text in AI video: trust Wan 3.0 text, or fix the image first?
Alibaba says Wan 3.0 renders text in 12 languages. For a price or legal line, edit the text in a still with Ideogram 4.5, then animate it: $0.70 per language.
- How long can an AI video be in 2026? Max duration by model
Vendor pages list 20 to 40 seconds per generation in 2026. Sume's catalog ids accept 15 s for most, 30 s for three, and 10 s for Omni Flash 1.1.
- Alibaba Live Avatar needs five H800 GPUs: or call a hosted clip API
Alibaba-Quark's open Live Avatar streams audio-driven video at 45 FPS on five H800 GPUs. Compare that hardware with a hosted still-plus-audio clip.
- Anam plans from $12 to $999: cost per live minute vs a rendered one
Anam's plan ladder works out to $0.12 to $0.24 per included live minute. A 60-second Sume avatar clip costs $11.04, so it wins above about 74 viewers.
Written by Sume