How to Use ImagineArt 2.0: 12 Aspect Ratios, 10 Credits Flat at 1K or 2K
A practical guide to ImagineArt 2.0 on VdoBloom: what reasoning-based generation changes, the real 1K and 2K pixel dimensions for all twelve aspect ratios, the flat 10-credit cost, and how to prompt it.
ImagineArt 2.0 is the reasoning-based member of the ImagineArt family on VdoBloom: it costs a flat 10 credits per image ($0.50 at the Lite rate) at either 1K or 2K, offers twelve aspect ratios — the widest selection of the three ImagineArt models, including 4:5 and 5:4 that neither 1.5 nor 1.5 Pro expose — and it is the one to use when a prompt describes a scene that has to make sense rather than just look pretty. It runs through Runware on model id imagineart:2.0@0 and is the newest of the three, added in April 2026.
What “reasoning-based” means in practice
It is not a marketing word for “better”. The practical difference shows up in prompts that contain relationships rather than a list of nouns: a subject holding a specific object in a specific hand, three items arranged in a stated order, one thing reflected in another, a sign that says something because of what is happening in the scene. Older text-to-image models render each noun beautifully and then put them in the wrong places. ImagineArt 2.0 is built to work through the arrangement first.
The second thing it is known for is colour. The model description in the roster calls out vibrant, true-to-life colour, and that matches how it behaves: saturated without going neon, and consistent across a set of images generated from similar prompts. That makes it a reasonable pick for a batch of images that must sit together in one deck, feed or catalogue.
Real specs and credit cost
Cost does not change with resolution. Both the 1K and 2K tiers bill the same 10 credits, which is unusual on the platform — most resolution-tiered models charge more for the larger output. Practically, that means there is no cost reason to generate at 1K. One credit is $0.05 at the Lite plan rate of $15 for 300 credits, so every image below is $0.50.
| Aspect ratio | 1K output | 2K output | Credits (either) | Cost |
|---|---|---|---|---|
| 1:1 | 1024 Ă— 1024 | 2048 Ă— 2048 | 10 | $0.50 |
| 16:9 | 1360 Ă— 768 | 2816 Ă— 1584 | 10 | $0.50 |
| 9:16 | 768 Ă— 1360 | 1584 Ă— 2816 | 10 | $0.50 |
| 4:3 | 1024 Ă— 768 | 2368 Ă— 1776 | 10 | $0.50 |
| 3:4 | 768 Ă— 1024 | 1776 Ă— 2368 | 10 | $0.50 |
| 3:2 | 1152 Ă— 768 | 2496 Ă— 1664 | 10 | $0.50 |
| 2:3 | 768 Ă— 1152 | 1664 Ă— 2496 | 10 | $0.50 |
| 21:9 | 1488 Ă— 640 | 3136 Ă— 1344 | 10 | $0.50 |
| 4:5 | 848 Ă— 1072 | 1856 Ă— 2320 | 10 | $0.50 |
| 5:4 | 1072 Ă— 848 | 2320 Ă— 1856 | 10 | $0.50 |
| 3:1 | 1680 Ă— 560 | 3552 Ă— 1184 | 10 | $0.50 |
| 1:3 | 560 Ă— 1680 | 1184 Ă— 3552 | 10 | $0.50 |
The 4:5 row is the one worth circling. 4:5 is the Instagram portrait frame, and 1856Ă—2320 at 2K is more than enough for a feed post with no cropping and no upscaling. The 5:4 row is its landscape twin for slide decks and blog headers.
What it does not do
ImagineArt 2.0 is text-to-image only. It appears in the generation model list and not in the editing list, so it accepts no reference images, no style transfer input and no masks. If your starting point is a photo, you want the image edit tab instead, where models accept between one and fourteen input images.
There is also no seed parameter in the generation call — the request carries the prompt and the resolved width and height and nothing else — so exact reproduction of an earlier image is not possible. Generate variations in one sitting and keep what you like. If a generation fails, the credits are refunded automatically.
How to prompt it well
Prompt ImagineArt 2.0 the way you would brief a person, because that is what the reasoning tuning rewards.
- State relationships explicitly. “A woman in a red coat standing to the left of a bicycle, her right hand on the handlebar” beats “woman, red coat, bicycle”. Left, right, behind, in front of, holding, resting on — those words do real work here.
- Give the scene a reason. A one-clause situation (“she has just stepped out of the rain”) produces coherent wet hair, damp pavement and reflected light without you listing each of those.
- Count things out loud. If you need exactly three bottles, say exactly three and say how they are arranged. Counting is where non-reasoning models fail most visibly.
- Name colours as materials. Instead of “vibrant blue”, say cobalt glazed ceramic or washed indigo denim. The colour tuning has more to grip.
- Always choose 2K. It is the same 10 credits. There is no reason to leave the model on 1K unless you specifically want a small file.
A worked example
Product flat-lay for a skincare launch, Instagram portrait. Choose 4:5 and 2K, giving 1856Ă—2320. Prompt: an overhead flat-lay on a pale limestone surface, exactly three frosted glass serum bottles arranged in a descending diagonal from top-left to bottom-right, a single eucalyptus sprig laid across the bottom-left corner, soft late-afternoon window light from the upper right casting long soft shadows to the lower left, muted sage and warm sand palette, generous empty space in the top-right third.
Every clause in that prompt is a constraint the reasoning step can satisfy: a count (exactly three), an arrangement (descending diagonal), a light direction and a matching shadow direction, and reserved negative space for copy. Cost: 10 credits, $0.50. A set of six variations for a feed test: 60 credits, $3.00.
When to pick a different model
| If you need | Better pick | Credits | Why |
|---|---|---|---|
| Native 4K in one pass | ImagineArt 1.5 Pro | 12 | Up to 6592 px wide; 2.0 stops at 2K |
| Hyper-real 2K for less | ImagineArt 1.5 | 6 | Same 2K ceiling, 4 credits cheaper |
| The cheapest image on the platform | Z-image Turbo | 2 | Fast turbo model for drafts |
| Best text and lettering | Seedream 4.5 Text-to-Image | 7 | Roster pick for typography, up to 4K |
| Non-Latin script text | Seedream 5.0 Pro | 10 / 19 | Multilingual, multi-reference fusion |
| To start from an existing photo | An edit-tab model | from 4 | 2.0 takes no image input at all |
The clean decision rule between the three ImagineArt models: 1.5 for cheap 2K realism, 2.0 when the composition has logic in it, 1.5 Pro when the file has to be enormous. Only 2.0 gives you 4:5 and 5:4.
New accounts get 10 credits at signup with no card, which is exactly one ImagineArt 2.0 image — enough to see whether the reasoning behaviour is what your prompts need. See what else that free balance stretches to in our free AI image generator breakdown, compare the whole roster in the multi-model image generator, check plan maths on the pricing page, or open image generation and pick ImagineArt 2.0 from the family selector.
Frequently asked questions
Does 2K cost more than 1K on ImagineArt 2.0?
No. Both resolutions bill a flat 10 credits, which is $0.50 at the Lite rate of $15 for 300 credits. Generate at 2K unless you have a specific reason to want a smaller file.
How many aspect ratios does ImagineArt 2.0 support?
Twelve: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 21:9, 4:5, 5:4, 3:1 and 1:3. The 4:5 and 5:4 options are unique to 2.0 within the ImagineArt family.
Can ImagineArt 2.0 edit an existing image?
No. It is text-to-image only and accepts no reference images. Use the edit tab for anything that starts from a photo.
Is ImagineArt 2.0 better than ImagineArt 1.5 Pro?
Not universally. 2.0 is newer, handles composition logic better and costs 2 credits less, but it stops at 2K. 1.5 Pro renders natively at 4K and beyond. Pick by output size first, then by prompt complexity.
What is the largest image ImagineArt 2.0 can make?
3552Ă—1184 on the 3:1 banner ratio, or 2048Ă—2048 for a square. Every listed ratio has both a 1K and a 2K variant.
Do paid plans remove the watermark?
Yes. Paid plans download watermark-free and include commercial rights; free-tier output is watermarked.
Ready to try it?
Create your first AI video in minutes — no credit card required.
Start Creating Free →