FREE
Loading…
The tier of ByteDance's 5.0 line that renders 4K — which Seedream 5.0 Pro cannot — at half the credits. Generate from a prompt, or attach a photo and change it with one instruction. No watermark.
Lite is not a cut-down Pro. It trades a little editing precision for the two things Pro does not offer: 4K, and half the price.
Native 4K output, not an upscale. Seedream 5.0 Pro has no 4K at all, so when the image has to be large this is the one of the two that can do it.
Same family, same one-endpoint design, half the credits. It is the sensible default for volume work and for iterating before committing.
Generate from a prompt or edit a photo with an instruction, in the same box — attach a reference to edit, omit it to generate.
The 5.0 line was built around photorealistic lighting and skin texture, and Lite inherits it.
Seedream 5.0 renders words inside an image far more reliably than most models, including non-Latin scripts.
New accounts get 5 free credits, and a Lite image costs exactly 5 — so the first one costs nothing.
One endpoint, both jobs. Attach an image and the prompt becomes an edit instruction; leave it out and the prompt generates from scratch.
This is the tier that renders 4K. Use 2K for screens, 4K when the image will be printed, cropped into, or shown large.
Roughly 20-40 seconds — 4K takes longer than 2K. Full resolution, no watermark, 5 credits.
4K takes longer and is worth it when you will print the image, crop deep into it, or show it large. For a website hero or a social post, 2K is already more than the surface needs.
With a reference image attached, describe only what should differ — "make it golden hour", "change the jacket to navy". Re-describing the whole scene invites the model to rebuild parts you liked.
At 5 credits Lite is the cheap place to find the composition. If the final frame needs the most precise possible edit and 2K is enough, re-run the winning prompt on Seedream 5.0 Pro.
Same family, same one-endpoint design. Two rows actually decide it: resolution and price.
| Feature | Seedream 5.0 Lite | Seedream 5.0 Pro |
|---|---|---|
| Max resolution | 4K | 2K |
| Smallest size | 2K | 1K |
| Credits per image | 5 | 10 |
| Reference images | One | One |
| Text-to-image + editing | Yes, one endpoint | Yes, one endpoint |
| Aspect ratio control | No — size only | No — size only |
| Pick it when | The image has to be big, or you are iterating | You want the most precise edit and 2K is enough |
Last updated: July 2026
Seedream 5.0 Lite is the cost-optimised tier of ByteDance's fifth-generation image model, running through the Volcano Ark official channel. Like Pro, one endpoint handles both text-to-image and instruction-based editing. Unlike Pro, it renders up to 4K.
Resolution and budget decide it. Lite renders 2K or 4K and costs 5 credits; Pro renders 1K or 2K and costs 10. If you need a 4K image, Lite is the only one of the two that can produce it. If you want the sharpest possible single-instruction edit and 2K is enough, Pro is the more precise instrument.
Yes — 4K is a native output size on Lite, not an upscale of a smaller render. That is the main reason to pick it over Pro, which has no 4K tier at all.
Both, through the same box. Attach a reference image and your prompt is treated as an edit instruction ("change the season to winter with snow"); leave the upload empty and the same prompt generates a new image.
No. Seedream 5.0 Lite takes one reference image. Multi-image fusion is a Seedream 4.5 capability; on sparkpix today, NanoBanana Pro is the model that accepts several images in one request.
Each image is 5 credits, whether you render 2K or 4K — half the price of Seedream 5.0 Pro. New accounts get 5 free credits on sign-up, so your first Lite image is free. Credit packs start at $9.99 for 350 credits, around 70 Lite images.
Yes. What you generate is yours to use commercially, at full resolution and with no watermark.
The flagship tier — the sharpest single-instruction editing, at 1K or 2K, for 10 credits.
Google Gemini 3 Pro Image — several reference images at once and the strongest in-image typography.
OpenAI’s image model — precise instruction-following and text-rich layouts.
Already have an image? Push it past 4K instead of regenerating it.
Describe an edit in plain text and the AI regenerates only what needs to change.
Generate from text at no cost with the SparkPix model — no credits required.