seedream 4.5
A ByteDance image model for reference-heavy edits, product preservation, typography, relighting, and composed advertising images, with support for up to 8 references and 4K output.
Key facts
| Developer | ByteDance |
|---|---|
| Official model ID | doubao-seedream-4-5-251128 |
| Availability on Banana Pie | Available now |
| Credits from | 25 credits / image |
| Reference images | Up to 8 |
Choose seedream 4.5 when an image job depends on several references, preserved product details, controlled scene changes, or output up to 4K. The supplied evidence shows broad scenario coverage rather than comparative quality, so it is a practical candidate for structured editing workflows but not a basis for assuming perfect text, anatomy, counting, or identity consistency.
Strengths
- Multi-reference capacity
Support for up to 8 reference images gives creators room to combine object, identity, garment, and scene guidance in one workflow.
- High-resolution output
The model supports output up to 4K for assets that need more detail or room for downstream layout work.
- Commercial image coverage
Published selections cover product lighting, advertising composition, product preservation, scene relighting, and cross-ratio outpainting.
- Text and layout workflows
The evaluation suite includes multilingual typography, structured UI graphics, and in-image text replacement, making it relevant to design-led prompts that mix imagery and copy.
Limitations
- No quality guarantee
The supplied selections document completed runs, not comparative wins or guaranteed prompt accuracy. Critical object details, anatomy, counts, and text still need review.
- Reference clarity matters
Up to 8 references gives the model more source material, but crowded or conflicting inputs can make it harder to determine which attributes should be preserved.
- Proofread generated text
Typography and text replacement were included in the evaluation suite, but the evidence does not establish error-free spelling across languages.
- Complex edits can take longer
The longest published selection in the supplied suite took 52.5s, so reference-heavy edits may require more waiting than simpler generations in this set.
Choose seedream 4.5 when
- You need to combine several visual references in one image workflow.
- You want to preserve a product while changing its setting, lighting, or composition.
- You are preparing advertising layouts, structured graphics, or images containing multilingual text.
- You need cross-ratio outpainting or coherent scene relighting.
- Your final asset benefits from output up to 4K.
Consider an alternative when
- You need guaranteed spelling, exact counting, or flawless attribute binding without review.
- You have conflicting references and cannot clearly indicate which source controls each detail.
- You need comparative proof that it outperforms another model; the supplied evidence does not provide head-to-head results.
- You are making a simple image where 4K output or multi-reference control adds no practical value.
Real-world results
Same fixed prompts we run on every model - one run each, published exactly as generated.
Prompt
A premium product photograph of exactly one transparent rectangular perfume bottle, half filled with amber liquid, standing upright on wet black stone. Light comes from the upper left, creating coherent refraction, a contact shadow, and one subtle reflection. No text, logo, plants, or extra objects.
Settings
- Resolution: 2K
- Aspect Ratio: 1:1
Prompt
Design a clean cream-colored vertical event poster. Show exactly four centered text lines and no other text: "MOONLIGHT MARKET", "月光市集", "18 OCT", and "RIVER HALL". Render "月光市集" in red and all other lines in black.
Settings
- Resolution: 2K
- Aspect Ratio: 2:3
Prompt
A photorealistic close-up of an adult potter shaping a clay bowl on a spinning wheel. Both hands are fully visible, each with five natural fingers touching the clay. No other people or hands. Soft window light.
Settings
- Resolution: 2K
- Aspect Ratio: 3:4
Prompt
On a matte gray table, exactly three red wooden cubes form a row on the left, exactly two blue glass spheres sit on the right, and one yellow ceramic mug stands centered behind them. The mug handle points right. No other objects or text.
Settings
- Resolution: 2K
- Aspect Ratio: 4:3
Prompt
Create a vertical social ad for a fictional running shoe named AERO. Keep the top 15 percent empty. Directly below it, place the headline "RUN LIGHT" in the upper-left. Show exactly one silver shoe in the lower-right, an orange trail curving from the bottom-left, and a round badge reading "42 KM". No other shoes or text.
Settings
- Resolution: 2K
- Aspect Ratio: 9:16
Prompt
Create a clean horizontal pricing comparison graphic titled "CHOOSE YOUR PLAN". Use exactly three equal columns labeled "STARTER", "PRO", and "TEAM". Under each column, show exactly three aligned rows labeled "PROJECTS", "STORAGE", and "SUPPORT", followed by one blue button labeled "SELECT". White background, dark navy text, no additional columns or text.
Settings
- Resolution: 2K
- Aspect Ratio: 16:9
Reference images
Prompt
Use Reference 1 as the base tabletop scene. Replace the red mug with the exact dark navy-blue mug from Reference 2 in the same position and orientation. Replace the folded white towel with the exact green-and-white striped napkin from Reference 3 in the same folded area. Replace the green pear with exactly one orange from Reference 4 in the same position. Replace the blue notebook with the exact mustard-yellow hardcover notebook from Reference 5 in the same position. Add exactly the two lemons from Reference 6 to the right of the mug. Preserve the wooden table, camera, framing, wood grain, lighting, shadows, and all other spatial relationships from Reference 1. Do not copy the white product backgrounds from References 2–6, duplicate any asset, or add other objects.
Settings
- Resolution: 2K
- Aspect Ratio: 4:3
Reference images
Prompt
Replace only the sign text with exactly "NIGHT OWL". Preserve the original font style, spacing, perspective, sign material, lighting, and everything else.
Settings
- Resolution: 2K
- Aspect Ratio: 3:2
Reference images
Prompt
Use the portrait in Reference 1 for the person and the isolated jacket in Reference 2 for the garment. Dress the person from Reference 1 in the exact jacket shown in Reference 2. Preserve the person's identity, face, expression, skin, hair, hands, pose, body proportions, background, framing, and lighting from Reference 1. Preserve the jacket's material, color, collar, buttons, pockets, and sleeve patch from Reference 2. Do not copy the ghost mannequin or white product background.
Settings
- Resolution: 2K
- Aspect Ratio: 3:4
Reference images
Prompt
Change the lighting to warm golden-hour sunlight entering from the left window. Do not move, add, remove, or redesign any object. Update highlights and shadows coherently.
Settings
- Resolution: 2K
- Aspect Ratio: 16:9
Reference images
Prompt
Expand the canvas to a 16:9 landscape by naturally continuing the beach, ocean, and sky on both sides. Keep the complete original image centered without cropping, stretching, letterboxing, or modifying it.
Settings
- Resolution: 2K
- Aspect Ratio: 16:9
Reference images
Prompt
Place the sneaker on a wet outdoor basketball court at dusk. Preserve the exact sneaker shape, black geometric side mark, materials, stitching, sole geometry, and camera angle. Add physically coherent contact, reflections, and dusk lighting.
Settings
- Resolution: 2K
- Aspect Ratio: 1:1
Credits & pricing
Generations start at 25 credits. In practice, that is the minimum shown for starting a seedream 4.5 job; confirm the displayed cost in Banana Pie before running higher-resolution or reference-heavy settings.
Frequently asked questions
How many reference images can I use?
It supports up to 8 reference images, making it suitable for edits and compositions that need to carry multiple objects, identities, garments, or visual cues into one result.
Can seedream 4.5 generate 4K images?
Yes. seedream 4.5 supports output up to 4K, which can help when the final image needs room for detailed product, advertising, or layout work.
Can it generate and replace text inside images?
The evaluation suite included multilingual typography and in-image text replacement. Those selections show that text-focused workflows were tested, but they do not establish error-free spelling or layout, so review important copy before publishing.
What kinds of reference-based edits were evaluated?
The suite covered six-reference object binding, identity and garment transfer, scene relighting, outpainting, and product preservation across a scene change. Use references that clearly show the details you need retained, then inspect the output for drift.
What does a generation cost?
A generation starts at 25 credits. The exact spend may depend on the options shown in the studio, so check the displayed cost before running a final 4K or reference-heavy job.
How we test & disclosure
Conclusions are based on a fixed prompt suite and its published selections covering product lighting, multilingual typography, anatomy, counting, advertising composition, structured graphics, reference binding, text replacement, relighting, outpainting, and product preservation.
Banana Pie sells paid access to this model alongside other models in one studio. Our verdicts come from tests run through the same pipeline our users get.