Seedream 4.5 is ByteDance's photoreal text-to-image model, sitting between Seedream 4.0 and the 5.0 line. It takes a prompt, an aspect ratio and a quality setting, and returns a single image.
From 21:9 through 9:16, chosen up front rather than cropped afterwards.
Signage and lettering in a scene come back legible more often than they used to.
Generated with Seedream 4.5 on this site on 2026-08-04, unretouched, at the basic quality setting. They were picked to show the three things this model is asked for most: skin and material texture, legible lettering inside a scene, and a flat-lay where every small object has to stay in focus.
Seedream 4.5 is a text-to-image model in ByteDance's Seedream family, positioned above 4.0 and below the 5.0 Lite and 5.0 Pro tiers. It takes three things: a written prompt, one of eight aspect ratios, and a quality setting of basic or high. There is no separate resolution ladder and no image count — one request returns one image.
What it is good at is the photoreal end of the range: material surfaces, skin, and the small print inside a scene. Text rendering is the part most worth knowing about, because it is where image models have historically failed in a way that is obvious to anyone looking; asking 4.5 for a sign, a label or a book spine returns something readable far more often than the generation before it did.
It is not an editor. Handing it an existing image to modify is a different model id on both shelves, and that path is not wired here.
Creative engine
One flat rate per generated image, independent of the aspect ratio you choose.
The mid tier of the Seedream family, and the cheapest one that does photoreal well.
Prompt, aspect ratio, quality. There is no seed, no batch and no resolution field.
Editing an existing image is a separate model and is not available on this page.
Pick 4.5 when the image has to look photographed and has to contain writing. That combination is narrower than it sounds and it is exactly where the cheaper tiers give up: Seedream 4.0 will produce a handsome image and turn the sign in the background into ornamental squiggles. Go up to 5.0 Pro when the subject is a person in motion, or when you need the extra fidelity enough to pay twice as much for the high quality setting. Stay on 4.0 when the picture is decorative and nobody will read anything inside it. The prompt transfers between all three, so testing down the ladder costs almost nothing.
Anything where the words in the frame are part of the point rather than set dressing.
Rust, condensation, worn leather, wet rope — texture is where photoreal is won or lost.
Generating at 21:9 or 9:16 directly beats cropping a square down to it afterwards.
What the model accepts, and what it deliberately does not.
One prompt in, one image out, with no source material required.
1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2 and 21:9.
Two settings rather than a resolution ladder; basic is what this page generates by default.
Words asked for inside the image come back legible more reliably than on 4.0.
Taken from the published KIE (kie.ai) schema for this model, which is the API it runs on here.
Four steps, and the third is the one that changes the result most.
Decide the shape before you write, because framing changes what belongs in the prompt.
Photoreal models respond to stated lighting far more than to adjectives like beautiful or stunning.
Put the words you want rendered in the prompt verbatim. Paraphrasing them is how they come back wrong.
Two images at 10 credits each is a cheaper way to find the shot than rewriting the prompt blind.
Jobs where the image has to survive being looked at closely.
Labels, bottles and boxes where the printed text has to read correctly.
People and places lit like a magazine shoot rather than an illustration.
Scenes built around lettering — shop fronts, menus, street signs.
Overhead arrangements where every object has to hold its own detail.
One credit pool covers Nano Banana images and Seedance and Veo video. The cost shows before every run, and failed jobs are not charged. Use a subscription for ongoing work, or a one-time pack when you just need to top up.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
One-time top-ups — buy extra credits any time you run low.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
What's included
Secure checkout by Stripe. Card details never touch our servers.
Charges appear as “KAVEL AI” on your card statement. You can cancel any time from Settings → Billing; cancellation takes effect at the end of the period you already paid for.
Powered by
Use one balance across every supported image and video model — ten of them today, listed below — and check the credit cost before each request.
Video models
Image models
Model credit guide
You're only charged for successful generations. The exact estimate in the generator varies by model, length, resolution, audio, and number of images.
| Type | Model | Credit cost |
|---|---|---|
| Video | Seedance 2.0 | 5s 720p image-to-video ≈ 188 credits; text-to-video ≈ 308 credits. Scales with resolution and length. |
| Video | Seedance 2 Fast | Faster and lower cost. 5s 720p text-to-video ≈ 248 credits. |
| Video | Seedance 2 Mini | The cheapest Seedance tier. 5s 720p ≈ 154 credits, 5s 480p ≈ 72 credits. |
| Video | Seedance 1.5 Pro | Audio doubles the rate. 5s 720p ≈ 27 credits silent, ≈ 53 with audio; 1080p ≈ 57 and ≈ 113. |
| Video | Veo 3.1 | Billed per video, not per second (Lite tier). About 45 credits at 720p and 53 at 1080p. |
| Video | Kling 3.0 | Audio raises the rate. 5s 720p ≈ 105 credits silent, ≈ 150 with audio; 1080p ≈ 135 and ≈ 203. |
| Video | MiniMax H3 | Fixed 2K, no resolution ladder. Priced per second — a 5s clip ≈ 158 credits. |
| Image | Nano Banana 2 | Generate or edit from text and images. About 8 credits per 1K image, 12 at 2K, 18 at 4K. |
| Image | Nano Banana Pro | Consistent run times across generations. About 12 credits per 1K or 2K image, 21 at 4K. |
| Image | Nano Banana 2 Lite | Faster, simpler variant. A flat 8 credits per image at every resolution. Start here to test. |
| Image | GPT Image 2 | The lowest-cost image model here. About 3 credits per 1K image, 6 at 2K, 12 at 4K. |
| Image | Seedream 5.0 Lite | Flat pricing — about 8 credits per image whether you generate or edit. |
Common questions about the mid tier of the Seedream family.
4.5 is the newer generation and is noticeably better at rendering text inside an image. 4.0 takes an image size and a 1K/2K/4K resolution instead of a quality setting, and costs a little less.
About 10 credits, flat, whichever aspect ratio you choose. Seedream 4.0 is about 8 and Seedream 5.0 Pro about 11 at its basic setting.
Not on this page. Seedream editing is a separate model id on the provider, and only the text-to-image path is wired here.
They are the two settings the API exposes. This page generates at basic; on the 5.0 Pro tier the same switch is also a price difference, which is why it is sent explicitly rather than left to a default.
Often, and far more often than earlier models, but not guaranteed. Quote the text verbatim and keep it short — long paragraphs inside an image are still unreliable on every model.
Credits are returned automatically, unless the request broke the content policy.
Pick a ratio, write the scene, quote any text you want in it, and run it.
Examples made on Kavel with this model.
One flat rate per image, whichever ratio and quality you pick.
