How to Use Luma UNI-1 Max: 8 Reference Images, Real Sizes and Credit Costs
A practical guide to Luma UNI-1 Max on VdoBloom: what it is best at, its fixed output dimensions, the exact 14 and 20 credit tiers, prompting rules and when a cheaper editor wins.
Luma UNI-1 Max is VdoBloom’s reference-image editor: you hand it up to 8 source images plus a written instruction, and it returns one edited image for 14 credits ($0.70) at the 1K tier or 20 credits ($1.00) at the 2K tier. It lives on the image editing tool rather than the text-to-image tool, because its whole reason to exist is working from pictures you already have — a product shot, a face, a room, a colour reference — and combining them into one coherent result.
What Luma UNI-1 Max is actually best at
Most image editors on VdoBloom take one image and change it. UNI-1 Max takes up to eight and reconciles them. That makes it the right tool for a specific family of jobs:
- Multi-reference composition. A person from image 1, an outfit from image 2, a location from image 3, and a lighting mood from image 4, merged into a single frame.
- Product placement. Put a packshot into a styled environment without rebuilding the product from a text prompt and hoping the label survives.
- Style transfer with an identity anchor. Keep the subject recognisable while a second reference supplies the grade, palette or art direction.
- Large-format output. Every aspect ratio it supports lands above 2000 pixels on the long edge, so the result is usable for a poster or a hero image, not just a thumbnail.
Its picker rating on VdoBloom is 4.3, which is honest: it is not the sharpest single-image retoucher in the lineup, and it is not the cheapest. It is the one that holds several references in its head at once.
The real specs
UNI-1 Max does not accept arbitrary pixel dimensions. It only accepts a fixed preset per aspect ratio, and the API rejects anything else, so the ratio you choose in the settings panel decides the exact output size. These are the presets VdoBloom sends:
| Aspect ratio | Output dimensions | 1K tier | 2K tier | Typical use |
|---|---|---|---|---|
| 1:1 | 2048 Ă— 2048 | 14 credits / $0.70 | 20 credits / $1.00 | Profile art, catalogue tiles |
| 16:9 | 2784 Ă— 1504 | 14 credits / $0.70 | 20 credits / $1.00 | Website hero, video thumbnail |
| 9:16 | 1504 Ă— 2784 | 14 credits / $0.70 | 20 credits / $1.00 | Story frame, reel cover |
| 4:3 | 2464 Ă— 1696 | 14 credits / $0.70 | 20 credits / $1.00 | Editorial, slide imagery |
| 3:4 | 1696 Ă— 2464 | 14 credits / $0.70 | 20 credits / $1.00 | Portrait, lookbook page |
| 3:2 | 2912 Ă— 1440 | 14 credits / $0.70 | 20 credits / $1.00 | Photographic landscape crop |
| 2:3 | 1440 Ă— 2912 | 14 credits / $0.70 | 20 credits / $1.00 | Poster, print-leaning portrait |
Other hard limits worth knowing before you start: the prompt is truncated at 3,000 characters, and reference images beyond the eighth are dropped rather than causing an error — so if you attach twelve, the last four are silently ignored. Order matters: put the image carrying the identity or the product first.
Dollar figures above use the Lite plan rate, where $15 buys 300 credits and one credit is therefore $0.05. Annual billing brings Lite to $10 per month, which quietly makes every number in that table about a third cheaper. Full plan details are on the pricing page, and new accounts start with 10 free credits and no card — just short of one 1K render, so spend those first on a cheaper editor.
How to prompt it well
UNI-1 Max responds to instructions that name its references explicitly. Vague prompts make it average the inputs into mush; specific prompts make it obey.
- Refer to images by position. Write the subject from the first image, the jacket from the second image. It tracks the ordering you uploaded.
- Say what must not change. One sentence like preserve the facial features, proportions and logo placement exactly does more work than three sentences of styling.
- Describe the light, not just the scene. Overcast window light from the left, soft falloff beats studio lighting.
- Pick the aspect ratio first. Because the output size is a fixed preset, you cannot crop back into detail later without losing resolution.
- Draft at 1K, finish at 2K. Iterate your prompt at 14 credits a shot and only spend the 20-credit tier once the composition is right.
A worked example
Say you are making a 9:16 ad frame for a skincare bottle. Upload four references in this order: the bottle packshot, a model portrait, a bathroom interior, and a colour-grade still you like. Then prompt something close to this:
The product from the first image held by the person from the second image, standing in the room from the third image. Preserve the bottle label text, shape and colour exactly, and preserve the person’s facial features. Soft morning window light from the left, gentle shadow falloff, muted warm palette matching the fourth image. Clean commercial photography, shallow depth of field.
Run it at 1K for 14 credits. If the label holds and the pose reads, re-run the same prompt and seed at 2K for 20 credits and you have a 1504 Ă— 2784 frame ready to drop into a story ad. Total spend to a finished asset: 34 credits, or $1.70 at the Lite rate.
When to pick a different model
UNI-1 Max is a mid-priced specialist, and there are jobs where something else in the image editor is the better call:
| Model | Credits per image | Pick it instead when |
|---|---|---|
| P-Image-Edit | 6 | You want fast, cheap iteration and up to 10 input images; volume matters more than polish |
| Seedream 5.0 Pro | 10 (Standard) / 19 (High) | You need coordinate or mask-precise local edits rather than a whole-frame recomposition |
| Kling IMAGE O1 | 14 | Same price, but you want tight control knobs and up to 10 references |
| Luma UNI-1 Max | 14 (1K) / 20 (2K) | You are blending several references into one composed frame at 2K-plus dimensions |
And if you have no source image at all, do not start here. Generate a base in the text-to-image tool first, then bring that output back into UNI-1 Max as reference number one. Editing from a real image is where this model earns its credits.
One note on people: if a reference image contains a real person, you need that person’s consent before you generate with it. VdoBloom’s flexible content tier covers swimwear, fitness, dance and fashion work that stricter tools refuse, but explicit and illegal content stays blocked on every model, UNI-1 Max included.
Frequently asked questions
How many reference images does Luma UNI-1 Max accept?
Eight. Anything beyond the eighth image is trimmed off before the request is sent, so attach your most important references first.
What is the difference between the 1K and 2K tiers?
It is the quality tier you are buying and the price step: 14 credits versus 20. The pixel dimensions themselves come from the aspect-ratio preset, so a 16:9 render is 2784 Ă— 1504 either way.
Can I use Luma UNI-1 Max for text-to-image?
On VdoBloom it is wired up as an editing model, so it expects at least one reference image. For prompt-only work, use a generation model and then refine the result here.
Do my images have a watermark?
Free accounts get watermarked output. Any paid plan downloads watermark-free with commercial rights, starting at $15 per month for 300 credits — about 21 UNI-1 Max renders at the 1K tier.
Do credits expire?
Subscription credits refresh with each billing cycle. One-time credit packs, from $2.49 for 75 credits, never expire.
Ready to try it?
Create your first AI video in minutes — no credit card required.
Start Creating Free →