It gets the intent
ByteDance presents Seedream 5.0 Lite as a model that grasps the intent behind an instruction, even a vague one, and follows complex requests: in its examples, up to nine subjects in a single image.
Image model · ByteDance Seed
Seedream v5.0 Lite — 2K images that follow your instructions, with or without references.
Seedream v5.0 Lite is the image model from ByteDance’s Seed team, introduced in February 2026 as a model that grasps the intent behind instructions, the way a human designer would. On Deepnia it creates 2K images in 7 aspect ratios from text, or edits and combines up to 10 reference images: a new backdrop, a style, a product or an outfit.
- Generation and editing
- Up to 10 references
- 7 aspect ratios
- 2K images
- 3,000-character prompt
The image tool opens with Seedream v5.0 Lite already selected. The exact cost shows before every creation; you sign up when you generate.
1:1
16:9
9:16
4:3
3:4
3:2
At a glance
- Maker
- ByteDance Seed
- Modes
- Generation, or editing with references
- References
- Up to 10 images
- Aspect ratios
- 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3 (1:1 by default)
- Size
- 2K, e.g. 2048×2048 in 1:1
- Prompt
- 3 to 3,000 characters
What Seedream v5.0 Lite does best
Edit without changing everything
Add a photo and describe the change, even in a few words. According to ByteDance, the model keeps the areas you did not ask to change more stable, for targeted edits.
Up to 10 reference images
Combine several images in one request: the outfit from one photo worn by the person in another, a product placed in a new setting, a logo applied to packaging. ByteDance documents this multi-image blending.
A style taken from one image
ByteDance highlights style transfer from a single reference image: the colors, light and brushwork of an artwork applied to your photo.
Small details done right
Letters, times, numbers and colors: ByteDance stresses the accuracy of these details, useful for a label, a schedule or a short tagline.
How to use Seedream v5.0 Lite on Deepnia
Open the image tool
The “Try Seedream v5.0 Lite” button opens Deepnia’s image generator with the model already selected.
Add references, or not
Without an image, you create from text. Use “Add an image” under “References” (up to 10) and the tool switches to “Edit mode”.
Describe the result
Write a prompt of 3 to 3,000 characters. With several references, name them in the order you added them: image 1, image 2…
Pick the ratio and run it
Pick the aspect ratio (1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3), check the cost shown, then click “Generate”, or “Edit” when you use references.
Seedream v5.0 Lite on Deepnia: what you can set
- Modes
- Generation from text; editing as soon as a reference image is added (“Edit mode”)
- References
- 1 to 10 images; JPG, PNG or WebP, up to 10 MB each
- Aspect ratios
- 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3 (1:1 by default)
- Target size per ratio
- 2048×2048 (1:1), 2560×1440 (16:9), 1440×2560 (9:16), 2560×1920 (4:3), 1920×2560 (3:4), 2560×1707 (3:2), 1707×2560 (2:3)
- Resolution
- 2K in each of the 7 aspect ratios
- Images per generation
- 1
- Prompt
- Required, 3 to 3,000 characters
- Text in the image
- Accurate letters, numbers and times according to ByteDance, for a label or a short tagline
Prompt ideas to try
Copy a prompt, open the tool and adapt it to your project.
Product shot for a craft lemonade
- Generation
- 3:4
Vertical advertising photo: a glass bottle of ginger lemonade on a light wooden table, surrounded by lemon slices, mint leaves and ice cubes. Late-morning natural light, blurred sunny terrace behind. High-end food photography, vivid natural colors.
New backdrop for a handmade soap
- Editing
- 1 reference image
- 1:1
Keep the soap and its packaging from the reference image exactly as they are, without changing anything on the product. Replace the background with a light wooden bathroom shelf, a white cotton towel, a green plant and soft light from the left. Natural shadow under the soap.
Poster for a music night
- Generation
- 2:3
Vertical poster for a rooftop music night at sunset: dancing silhouettes, string lights, orange and purple palette. At the top, in large white letters: "SUNSET SESSION". At the bottom, in small letters: "SAT 8 PM". Clean, modern graphic style.
Style transfer for a bookshop
- Editing
- 2 reference images
- 4:3
Turn the photo of the bookshop (image 1) into an illustration in the style of the painting in image 2: same colors, same brush texture, same outlines. Keep the composition, the shop window and the passers-by from image 1.
Virtual try-on for a fashion brand
- Editing
- 2 reference images
- 3:4
Dress the woman in image 1 in the dress from image 2, faithfully keeping the fabric’s patterns, colors and cut. Keep her face, hairstyle and pose. Plain beige studio background, soft fashion lighting.
Team photo for a startup in Nairobi
- Generation
- 16:9
Realistic team photo of six young professionals in a bright coworking space in Nairobi: two standing at a whiteboard, two sitting around a laptop, one laughing with a coffee, one taking notes. Modern outfits, natural light, warm and professional mood.
Tips for better results
- With several references, give each one a role in the order you added them: “the dress from image 2 on the woman in image 1”.
- When editing, also say what must stay the same (product, face, framing): the instruction gets clearer.
- To borrow a style, one style image is enough: ByteDance highlights style transfer from a single reference.
- Put the text to display in quotation marks, keep it short and check every letter before you publish.
- Pick the aspect ratio for where the image will go: 3:4 for a product post, 9:16 for a story, 16:9 for a banner or a website.
Good to know
- As soon as you add a reference image, the tool switches to “Edit mode”; remove the references to go back to creating from text alone.
- To refine an image you already created, add it as a reference from your Deepnia library, then describe the change you want.
- For a carousel, a storyboard or a set of related visuals, switch to Seedream v5.0 Lite Sequential: up to 15 consistent images in a single request.
- For virtual try-ons or portrait edits, use photos you have the rights to, with the consent of the people pictured.
Seedream v5.0 Lite or another model?
Deepnia’s other image models, to pick the right one for your project.
| Model | Resolution | Aspect ratios | References | Best for |
|---|---|---|---|---|
| Seedream v5.0 LiteThis model | Automatic | 7 ratios | Up to 10 | 2K images that follow your instructions, with or without references. |
| Seedream v5.0 Lite Sequential | Automatic | 7 ratios | Up to 10 | A consistent series of 1 to 15 images from a single prompt. |
| Nano Banana 2 | 1K–4K | 14 ratios | Up to 14 | Images in 1K, 2K or 4K, as a PNG or JPG file. |
| GPT Image 2 | 1K–4K | 8 ratios | Up to 16 | Text-rich images: infographics, posters and graphic design. |
| Nano Banana 2 Lite | 1K | 14 ratios | Up to 10 | The default model, with very wide or very tall ratios such as 21:9 or 4:1. |
| Grok Image | Automatic | 5 ratios | Up to 1 | Editing a single photo while keeping its original ratio. |
Frequently asked questions
What resolution are Seedream v5.0 Lite images on Deepnia?
2K images in each of the 7 aspect ratios: Deepnia targets, for example, 2048×2048 in 1:1 and 1440×2560 in 9:16.
How many reference images can I add?
Up to 10 (JPG, PNG or WebP, up to 10 MB each), from your device or your Deepnia library. In the prompt, refer to them in the order you added them: image 1, image 2…
Can I edit an existing photo?
Yes: as soon as you add a reference image, the tool switches to “Edit mode”. ByteDance says the model understands short or vague instructions and keeps the unedited areas stable.
Can Seedream v5.0 Lite write text in the image?
Yes: ByteDance highlights accurate letters, numbers and times. Put the text in quotation marks, keep it short and check it. For a text-heavy infographic, GPT Image 2 is a better fit.
How do I write a good prompt for Seedream v5.0 Lite?
Describe the subject, the setting, the light and the style in plain sentences. When editing, say what changes and what stays the same; with several references, give each one a role in the order you added them, for example “the dress from image 2 on the woman in image 1”.
Which aspect ratios are available?
1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3; 1:1 is selected by default, and each ratio has its own 2K size.
How do I keep the same product or character across several images?
Add the same reference photo to each creation, describe the product or character with the same words and say what must stay identical. For a whole series in a single request, choose Seedream v5.0 Lite Sequential.
What is the difference with Seedream v5.0 Lite Sequential?
It is the same ByteDance model for two uses: Seedream v5.0 Lite creates 1 image per generation, the Sequential version a consistent series of 1 to 15 images in a single request.
Ready to try Seedream v5.0 Lite?
The image tool opens with Seedream v5.0 Lite already selected. The exact cost shows before every creation; you sign up when you generate.