Ideogram 4.5 vs Seedream 4.5 for image edits: limits and price on Sume
Both edit via POST /v1/images. Ideogram 4.5 takes 5 references, priced by quality; Seedream 4.5 takes 10 at one rate. Catalog facts, read 2026-10-05.
Pick Ideogram 4.5 when the edit changes text and must leave the rest of the picture alone, and Seedream 4.5 when you want up to 10 references in one call at a single flat rate. On Sume, both go through POST /v1/images: ideogram/ideogram-v4.5 costs $0.0375, $0.075 or $0.275 per edit by quality, and bytedance-seed/seedream-4.5 costs $0.05.
The numbers below come from Sume's catalog and docs. Nothing here is a quality benchmark: Ideogram's own claim is that 4.5 is "the most precise edit model" (Ideogram on X, read 2026-10-05), and no equivalent claim is in this comparison for Seedream. The only fair test is the one at the end of this post.
How do the two compare field by field?
Prices are fal list prices times Sume's 1.25. The Image API docs give Ideogram's list as $0.03, $0.06 and $0.22 by quality; Seedream 4.5's list is $0.04 in Sume's pricing table.
| Field | Ideogram 4.5 | Seedream 4.5 |
|---|---|---|
| Public id | ideogram/ideogram-v4.5 | bytedance-seed/seedream-4.5 |
| Price per image at Sume | $0.0375 low, $0.075 medium, $0.275 high | $0.05 |
| References per request | 5 (first is edited, up to 4 more are references) | up to 10 |
| Quality parameter | low, medium (default), high | none |
| Resolution tiers | 1K, 2K | none listed |
| Aspect ratios | 15, from 1:3 to 3:1 | 9, from 2:3 to 3:2 and 16:9, 9:16 |
output_format | not accepted (provider picks) | png, jpeg, webp |
mask_url | 400 unsupported_parameter | 400 unsupported_parameter |
When does the reference limit decide it?
An Ideogram edit uses the first image as the picture to change and takes up to four more as references: five in all. If you need a product shot, a logo, a swatch, a style sheet and a pose reference in one call, that is five slots and Ideogram fits; a sixth does not. Seedream accepts up to 10, which suits composites where you feed several product angles at once. How many reference images does each Sume image model take lists every row.
Neither model takes a mask on Sume. mask_url is live only on ChatGPT Image 2.5, so a precise region edit needs that model or a prompt that names the region.
When does price decide it?
At low, Ideogram is cheaper than Seedream. At medium it is 50 percent more, and at high it is 5.5 times more. For a text-only change like a headline swap, low is worth trying first, which is why the table below prices ten edits at each tier.
| Model and quality | Per edit | 10 edits | 100 edits |
|---|---|---|---|
| Ideogram 4.5 low | $0.0375 | $0.375 | $3.75 |
| Ideogram 4.5 medium | $0.075 | $0.75 | $7.50 |
| Ideogram 4.5 high | $0.275 | $2.75 | $27.50 |
| Seedream 4.5 | $0.05 | $0.50 | $5.00 |
What about the shape of the result?
An Ideogram edit without aspect_ratio keeps the source geometry, per the Image API docs, so a 4:5 poster stays 4:5. Seedream's ratio list is shorter and has no 1:3 or 3:1, so a tall or wide banner edit is an Ideogram-only job among these two. Check the aspect_ratio descriptor in GET /v1/images/models/{id}/endpoints before you build around either list.
A fair test in 30 minutes
Take 12 real edits from your own queue: four text swaps, four recolors, four object changes. Run each on Ideogram at medium and on Seedream, and keep the prompts word for word identical. Hide the model names and have two people mark each result pass or fail on one question: did anything change that you did not ask for? The blind A/B sheet recipe builds the sheet. Total spend is 12 x $0.075 + 12 x $0.05 = $1.50, and the answer is yours, not a vendor's.
A small test before the full run
Before you scale this up, run it on two or three real files first and write down what you saw. A small test at low quality costs cents on Sume ($0.0375 per Ideogram 4.5 edit), and it tells you whether your prompt, your source files and your review step are ready.
Keep the originals untouched, name every output after its source and its prompt, and store the job id with each result. If a result is wrong later, you can find the exact request, fix the prompt and re-run only that item with a new Idempotency-Key.
Sources
Related posts
More in Comparisons
- Is Grok Imagine cheaper than GPT Image 2.5? Only versus high
On Sume, Grok Imagine costs $0.025 per image. GPT Image 2.5 is cheaper at low and medium, and Grok is cheaper only at high, xhigh and max. Per-tier table.
- Kaltura conversational avatars vs Sume clips for a training portal
Kaltura launched real-time avatars with an SDK on 2026-03-12. For a training portal where lessons are fixed, Sume renders the avatar clips instead.
- Kling 3 two 15 s jobs make a 30 s clip: $4.30 silent, $6.40 with sound
Kling 4.0 allows 30 s in one clip. On Sume, Kling 3 stops at 15 s, so run two jobs and join them: $4.30 silent or $6.40 with audio, including the join.
- Kling 4.0's 15 references vs Sume: which video model takes the most
Kling 4.0 lists up to 15 references: 10 images and 5 videos. On Sume, Wan 3.0 takes 10 images, 5 videos and 5 audio clips; Omni 10 images and 3 short videos.
Written by Sume