AI Image Generator
Models
Nano

Google's cutting-edge image generation model

3 credits
Fast
High Quality
Gpt

OpenAI's GPT-4o image generation model

3 credits
High Quality
Text & Image
Seedream

ByteDance image generation and reference-based editing in 2K, 3K, or 4K

6 credits
2K/3K/4K
Text & Image
Flux

Flux Kontext standard model with balanced performance

4 credits
Fast
High Quality

Powered by ByteDance Seedream 4.5 API

Public Visibility

When this option is enabled, the output image may be selected by AI Generate and published to the Explore.

Notification
This is a notification message
Sample Images
ByteDance image model on Lovart ME

Seedream 4.5 AI Image Generator Build a new visual or direct a reference-based edit

Start with a written brief or upload reference images, choose one of eight aspect ratios, and render at 2K or 4K. Each output costs 7 credits in this integration.

4K
Maximum output
Choose 2K or 4K
8
Aspect ratios
Square, portrait, landscape, and ultrawide
4
Reference slots
A Lovart ME workflow limit
7
Credits per image
Integer price for either mode

Use Seedream 4.5 for controlled visual production

The model can begin from text or use several images as visual direction. Good results still depend on assigning each reference a clear role and separating fixed details from requested changes.

Text-led image generation

Define the deliverable, subject, composition, lighting, palette, and exclusions. A concrete brief gives you visible criteria for reviewing the first output.

Reference-aware editing

Keep selected traits such as facial features, product geometry, lighting, or color treatment while asking for a specific scene, material, or layout change.

Multi-image composition

Use different references for the subject, environment, palette, or design language. State which source controls each part so the instructions do not compete.

Layout and text direction

Seedream 4.5 is designed for typography and dense-text layouts, but important copy still needs review. Specify exact wording, hierarchy, placement, and clean space.

Choose 2K or 4K for the job in front of you

The API maps Basic quality to 2K and High quality to 4K. Both options cost 7 credits here, so choose based on review and delivery needs.

2K for prompt and layout decisions

Use 2K while comparing composition, reference roles, copy space, and art direction. It is the practical default for iteration and costs 7 credits.

4K for closer inspection

Choose 4K when the direction is stable and you need more pixels for edge review, crop flexibility, or a larger deliverable. It also costs 7 credits.

Match the eight ratios to the final placement

01

Square and balanced

Use 1:1 for catalog tiles, profile posts, centered products, and compact campaign assets.

02

Portrait and mobile

Use 2:3, 3:4, or 9:16 for covers, posters, people, stories, and full-screen mobile placements.

03

Landscape and editorial

Use 3:2, 4:3, or 16:9 for articles, presentation visuals, web headers, and scene-based compositions.

04

Ultrawide banners

Use 21:9 for wide headers and backdrops. Tell the model where the subject sits and which side must remain clear for copy.

Seedream 4.5 prompt templates for real design work

Replace the bracketed fields, remove irrelevant instructions, and identify references by order. For edits, list what stays fixed before describing what changes.

Controlled product edit

Change one product property while preserving its construction.

Prompt Edit reference 1. Preserve the exact silhouette, dimensions, seams, controls, label position, and camera angle. Change only [material or color] to [new treatment]. Match realistic reflections and contact shadows. Keep the background [description]. No extra objects or redesigned details.

Working note: Name every feature that must remain stable; do not rely on a general instruction such as keep it similar.

Multi-reference composite

Combine separate visual sources with explicit responsibilities.

Prompt Create a [ratio] scene using the subject identity and wardrobe from reference 1, the environment from reference 2, and the palette and lighting mood from reference 3. Preserve the subject's facial features. Match perspective, scale, reflected light, and contact shadows. Do not copy text or logos from any reference.

Working note: Remove any reference that does not control a specific visible decision.

Editorial poster system

Create a structured poster direction with dense information zones.

Prompt Design a [ratio] editorial poster for [topic]. Use this exact title: [title]. Add these short supporting lines: [copy]. Hierarchy: title first, subject second, details third. Grid: [description]. Palette: [colors]. Keep all text horizontal, legible, and separated from the subject. No watermark or filler copy.

Working note: Keep generated copy short and inspect it at full size before using the asset.

How to use Seedream 4.5 on Lovart ME

Treat the generator as a production brief. Change one decision at a time so you can tell which instruction improved the result.

  1. 01

    Choose text or image mode

    Use Text to image for a new concept. Choose Image to image when a subject, product, composition, palette, or other visible property must come from a source.

  2. 02

    Write the brief and assign references

    State the deliverable first, then subject, setting, composition, lighting, copy, and exclusions. Give each uploaded image one named role.

  3. 03

    Set ratio and resolution

    Choose a ratio for the final publishing slot. Start at 2K for working decisions or select 4K when you need a larger file for closer review.

  4. 04

    Generate and inspect

    Review identity, hands, product geometry, small text, edges, shadows, and unwanted additions. Tighten specific constraints before generating again.

Seedream 4.5 questions

These answers follow the published Kie.ai parameters and the controls available in this Lovart ME workflow.

What is Seedream 4.5?

Seedream 4.5 is a ByteDance image model for text generation and reference-based image editing. This page connects the separate Kie.ai text-to-image and edit endpoints in one workspace.

Can Seedream 4.5 edit an existing image?

Yes. Switch to Image to image, upload at least one reference, and describe the requested changes along with the details that must remain fixed.

How many credits does one image cost?

Both text generation and image editing cost 7 integer credits per output. The supplied upstream listing shows 6.5 credits, which this project rounds up as requested.

Which resolutions are available?

The documented API provides Basic quality at 2K and High quality at 4K. Both resolution options cost 7 credits in this integration.

Which aspect ratios can I use?

The available ratios are 1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, and 21:9.

Which reference files are accepted?

The edit API accepts JPEG, PNG, and WebP images up to 10MB each. This Lovart ME workflow exposes four reference slots.

Will generated text always be correct?

No image model should be treated as a final proofing tool. Keep copy short, request exact wording and layout, then inspect every character or replace critical text in a design editor.

Can I use the result commercially?

Usage depends on Lovart ME terms, your rights to uploaded references, and applicable law. Confirm that you can use the source material and review current terms before publishing commercially.