Seedream 5.0 Pro is a third-party image generation model from ByteDance, now available on Runway. It follows detailed prompts closely, renders in-image text and graphics, can follow a sketch reference with high fidelity, and can combine several reference images into a single cohesive output. Use it with a text prompt alone, or guide it further with reference images or sketch-based edits.

This article covers how to access the Seedream 5.0 Pro in Custom mode, prompt and input best practices, available settings, and what to expect from your generations.

Spec information

Supported inputs	Text prompt, reference images (optional)
Max reference images	10
Supported aspect ratios	1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3
Credit cost by resolution
1K	5 credits per image
2K	10 credits per image

Step 1 — Accessing Seedream 5.0 Pro

Seedream 5.0 Pro is available in the Custom creation mode in Runway.

Select Custom in the left sidebar
At the top of the interface, confirm Image is selected — Seedream 5.0 Pro is an image model and will only appear when Image mode is active
Open the model picker and select Seedream 5.0 Pro

Step 2 — Adding the inputs

Seedream 5.0 Pro accepts a text prompt and optional visual inputs.

Prompt

Describe the image you want to generate.The model follows detailed, multi-step instructions closely, including the text you want rendered in the image.

Image references (optional)

Combine elements from multiple reference images into a single output — for example, a subject from one image, an object from a second, and an accessory from a third. Supports up to 10 reference images.

Sketch or marked-region edits (optional)

Guide edits by drawing directly on an image — circle or mark an area and describe the change, or provide a rough sketch for the model to render realistically.

Step 3 — Generating with Seedream 5.0 Pro

Before generating, review your settings:

Output count — set to 4 by default
Resolution — choose between 1K and 2K
Aspect ratio — set to 16:9 by default

Credit costs are tied to the selected Resolution.

Once your prompt and settings are ready, select Generate.

If you are using Agent, you can type directly in your prompt the resolution and aspect ratio you'd like for your generated image.

Examples outputs

Next steps

You've used Seedream 5.0 Pro to generate images with strong character consistency and in-image text. From here, you might explore apps to further work with your new images:

Story Panels — Expand the world and story of your image with new perspectives
Multi-Shot Video — Turn a product or campaign image into a multi-shot video sequence

Type	Referenc images	Prompt
Multi-image references		A woman in her 60s with silver locs, wearing a canvas apron, standing behind a wooden market stall. She's holding the ceramic honey jar from the second reference image and wearing the woven sun hat from the third. Morning market setting, string lights overhead, wooden crates of produce softly out of focus behind her, warm early light, documentary-style commercial photography, natural unposed expression.
Text and graphic rendering	None	A clean, single-panel infographic titled "Neighborhood Coffee Co: From Bean to Cup" with a horizontal five-step process flow: "1. Sourced" "2. Roasted" "3. Ground" "4. Brewed" "5. Served," each with a simple line-icon above it. Warm cream background, terracotta and forest-green accent colors, modern sans-serif typography, small-batch coffee shop branding aesthetic.
Sketch references		Edit the image based on the sketching. Add a giant garden gnome sitting on the house in the upper-left marked area, move the cow from the fiel to the top of the red barn, replace the stick figure with a robot farmer, and write Odd Day in the clouds in the top right area. Remove all the read and white sketched lines. Keep the composition unchanged. Make the newly added objects blend naturally into the original scene.
Multi-subject positioning		Place only the people in @Image 1 in the corresponding seat in @Image 2 marked with the matching circle color.