How to Use Kling IMAGE O1: 10 References, 2K Output, 14 Credits
A practical guide to Kling IMAGE O1 on VdoBloom: its real 1K and 2K dimensions, the 10-reference limit, the flat 14-credit price, prompting rules and when a cheaper editor wins.
Kling IMAGE O1 is VdoBloom’s high-control image model: it accepts up to 10 reference images plus a text instruction and returns an edited image at 1K or 2K for a flat 14 credits — about $0.70 — with no surcharge for the higher resolution and no surcharge for extra reference images. It runs on the klingai:kling-image@o1 endpoint, it is tagged MODERATE on VdoBloom’s content-filter scale, and it carries a 4.0 community rating in the picker. This guide covers exactly what it can do, the numbers it actually runs on, how to prompt it, and the three situations where a different model on VdoBloom is the smarter buy.
What Kling IMAGE O1 is best at
The single thing Kling IMAGE O1 does better than most of the cheap editors is hold a character steady while you change everything around them. Because it takes a stack of reference images rather than one, you can feed it three or four photos of the same face from different angles and then ask for a new outfit, a new background, or a new pose, and the identity survives the edit. That is the whole reason it exists in the roster alongside faster and cheaper editors.
Concretely, it is a good fit for:
- Consistent character sheets. Same person, six different scenes, one reference stack.
- Product-plus-model composites. Reference one product shot and one model shot, then instruct the combination.
- Precise local edits. Change the jacket, keep the face, keep the lighting.
- Pose variations. Kling IMAGE O1 is one of the models wired into VdoBloom’s pose-variation generator, which fans a single reference out into a batch of new poses.
It is not the fastest model on the platform and it is not the cheapest. It is the one you reach for when a cheaper edit has already failed on identity drift.
Real specs: aspect ratios and delivered pixels
Two resolution presets are available, 1K and 2K, across eight aspect ratios. These are the exact dimensions the model renders in text-to-image mode — they come straight from the dimension map the service uses, not from a marketing page.
| Aspect ratio | 1K output | 2K output | Typical use |
|---|---|---|---|
| 1:1 | 1024 Ă— 1024 | 2048 Ă— 2048 | Avatars, catalogue tiles |
| 3:2 | 1248 Ă— 832 | 2496 Ă— 1664 | Editorial stills |
| 2:3 | 832 Ă— 1248 | 1664 Ă— 2496 | Portrait, print |
| 4:3 | 1168 Ă— 880 | 2336 Ă— 1760 | Classic landscape |
| 3:4 | 880 Ă— 1168 | 1760 Ă— 2336 | Lookbook portrait |
| 9:16 | 768 Ă— 1360 | 1536 Ă— 2720 | Reels and Shorts frames |
| 16:9 | 1360 Ă— 768 | 2720 Ă— 1536 | Thumbnails, banners |
| 21:9 | 1552 Ă— 656 | 3104 Ă— 1312 | Ultra-wide hero images |
Other hard limits worth knowing before you start: up to 10 reference images per request, and a prompt of at least 3 and at most 2500 characters. VdoBloom rejects a prompt under three characters with a 400 before it ever costs you a credit, and it truncates at 2500. A seed field is available, so an output you like can be reproduced exactly.
One quirk that catches people out: in editing mode the model is driven by the resolution preset and the references, not by width and height. That means the aspect ratio you pick is authoritative in text-to-image mode, while an edit tends to follow the shape of what you fed it. If you need a specific crop out of an edit, get the reference into that crop first.
What Kling IMAGE O1 costs
One credit on VdoBloom is $0.05 at the Lite price of $15 for 300 credits. Kling IMAGE O1 is flat-rate, which makes the maths unusually simple.
| What you run | Credits | Dollar cost | What you get |
|---|---|---|---|
| 1K edit, 1 reference | 14 | $0.70 | One image up to 1552 px on the long edge |
| 2K edit, 1 reference | 14 | $0.70 | Same price, double the linear resolution |
| 2K edit, 10 references | 14 | $0.70 | No per-reference charge |
| Text-to-image, no reference | 14 | $0.70 | Dimensions from the table above |
| Pose variations, batch of 4 | 56 | $2.80 | Four posed variants of one subject |
Because the 2K tier is free, there is almost never a reason to run this model at 1K unless you are deliberately keeping file sizes down. Full plan pricing, including the annual rate and the credit packs that never expire, is on the pricing page. New accounts start with 10 free credits, which is not quite enough for one Kling IMAGE O1 render — use those on a cheaper editor first and save this model for the job that needs it.
How to prompt it well
Kling IMAGE O1 rewards instruction-style prompts over tag soup. Four rules that hold up in practice:
- Name what must not change. Write the preservation clause explicitly: keep the face, hair, skin tone and body proportions from the references unchanged. The model has the capacity to hold identity; it needs to be told that is the priority.
- Order your references deliberately. The first reference carries the most weight. Put the subject first and the style or product references after it.
- One structural change per render. Outfit and background and pose in a single instruction produces mush. Chain three edits at 14 credits each instead, and keep the seed so you can step back.
- Describe light, not adjectives. Soft window light from camera left, with a faint falloff on the right cheek beats cinematic.
A worked example
Say you have three phone photos of the same model and you need a 9:16 lookbook frame.
Upload all three on the image editor, pick Kling IMAGE O1, set resolution to 2K, and prompt: Using the person in the reference images, place her in a clean studio with a warm grey seamless backdrop. Replace the outfit with a tailored cream linen suit. Keep her face, hairline, skin tone and body proportions exactly as in the references. Soft key light from camera left, gentle shadow falloff, editorial lookbook framing.
That is 14 credits, $0.70. Note the seed on the result. For the second frame, reuse the same references and change only the sentence about the outfit. Four frames costs 56 credits, and the face is the same face in all of them — which is the whole point. If you need to start from nothing instead of a photo, the same model is reachable from the image generator with no reference attached.
When to pick a different model instead
Fourteen credits is mid-table on VdoBloom, and plenty of jobs do not need what this model offers.
| If you need | Better pick | Credits |
|---|---|---|
| Cheap bulk edits | P-Image-Edit | 6 |
| Cheapest usable editor | GPT Image 2 at 1K | 3 |
| 4K output | Nano Banana Pro 4K | 24 |
| Flexible content filter | Seedream 5.0 Pro Standard | 10 |
| Design-grade typography | Riverflow 2.0 Pro | 18 |
Kling IMAGE O1 tops out at 2K, so if you need a genuine 4K deliverable you must go elsewhere. It is also tagged MODERATE, which means it declines more swimwear and dance material than the models on the flexible tier; for that work the Seedream and WAN families are the right call. Explicit and illegal content stays blocked on every model.
If you are editing a photo of a real person, you need that person’s consent before you generate anything with their face.
Disclosure: VdoBloom is our platform. Competitor details were checked against their own site and we name the competitor as the better pick wherever that is true. Kuaishou’s own Kling app sells IMAGE O1 directly, and if Kling models are the only thing you will ever use, buying from Kuaishou is cheaper per image than buying credits that also cover 105+ other models. Our Kling comparison page lays that trade-off out honestly. VdoBloom’s case is the one credit balance that also reaches the video side — the spec sheet and sample output for this model live on its model page, and its video sibling is on the Kling V2.5 Turbo page.
Frequently asked questions
Does 2K cost more than 1K on Kling IMAGE O1?
No. Both resolutions cost a flat 14 credits, about $0.70. Run 2K by default.
How many reference images can I upload?
Up to 10. The price does not change with the number of references, and the first reference carries the most weight in the result.
Can Kling IMAGE O1 generate an image without any reference?
Yes. With no reference attached it runs as text-to-image and uses the aspect-ratio dimension table above. With one or more references attached it runs as an edit.
Is there a watermark on the output?
Paid plans download watermark-free with commercial rights. Free accounts get a watermark.
Why was my prompt rejected before generating?
Prompts shorter than three characters are rejected outright, and anything past 2500 characters is truncated. Neither costs you a credit.
Can I reproduce the same image twice?
Yes. Keep the seed from a result you like and pass it back with the same references and prompt.
Ready to try it?
Create your first AI video in minutes — no credit card required.
Start Creating Free →