flux-2/turbo/text-to-image
Create detailed visual assets from prompts with scalable, high-speed precision
Seedream 4.5 belongs to the newest generation of image models by Bytedance. Compared to previous iterations, Seedream 4.5 offers significantly improved prompt adherence and clearer details in edited regions. While this page focuses on the Edit mode, users looking to generate images from scratch should use the Seedream 4.5 Text-to-Image model.
On RunComfy, Seedream 4.5 is hosted as a managed high-performance service, exposed in two complementary ways:
Upload your image, type your edit instructions, and run jobs directly in your browser. Ideal for exploring prompts and testing how Seedream 4.5 interprets specific edits without writing code.
Developers can integrate these capabilities directly into their applications using the Playground API. This provides a production-ready endpoint to call Seedream 4.5 programmatically.
👉 Get started with the API here: Seedream 4.5 Edit API
In all cases, inference runs on RunComfy’s cloud GPUs—no local hardware, drivers, or downloads needed.
Seedream 4.5 Edit requires specific inputs to ensure the best editing results. Below are the parameters configured for this model.
| Parameter | Type | Description |
|---|---|---|
| prompt | string | The instruction for editing the image. Note: The recommended length is no more than 600 English words. If the prompt is too long, Seedream 4.5 may ignore details and concentrate only on main points, resulting in missing elements. |
| image_urls | string | Required. The source image(s) you want to edit. Seedream 4.5 uses this as the baseline for modifications. |
| Parameter | Type | Options | Description |
|---|---|---|---|
| resolution | enum | 2048x2048 (1:1), 2304x1728 (4:3), 1728x2304 (3:4), 2560x1440 (16:9), 1440x2560 (9:16), 2496x1664 (3:2), 1664x2496 (2:3), 3024x1296 (21:9) | Defines the output quality and aspect ratio. Default is 2048x2048 (1:1). Choosing higher resolutions like 4K allows Seedream 4.5 to generate ultra-detailed textures. |
Keep your prompt concise. If you want to change lighting, specify "Golden hour" or "Neon lights" clearly. Ensure your source image resolution matches the target resolution parameter for the best pixel-perfect mapping.
When asking Seedream 4.5 to change major elements (e.g., "Replace the field with an ocean"), ensure the prompt describes the new scene fully. Remember the 600-word limit to prevent the model from losing focus.
For social media, use 1728x2304 (3:4) or 1440x2560 (9:16). For desktop wallpapers or high-quality prints, select 2560x1440 (16:9) or 4K.
Seedream 4.5 targets professional-grade outputs with a focus on photorealism and texture accuracy.
Seedream 4.5 Edit shines across entertainment, film creation, and e-commerce workflows.
Seedream 4.5 delivers an overall upgrade compared to Seedream 4.0, especially in editing stability, small-text clarity, and portrait rendering.
In short, Seedream 4.5 builds on Seedream 4.0 with more reliable edits, clearer details, and higher-end artistic quality, helping creators turn precise ideas into visually refined images.
Explore the research and full capabilities of the Seedream family at the official site:
Create detailed visual assets from prompts with scalable, high-speed precision
Nail the art of text and vector imagery.
Create lifelike visuals and illustrations from text with flexible design control.
Turn sketches into precise 2K-4K visuals with smart correction and seamless creative control.
Multi-angle image editing with precision control and seamless visual consistency
Generate images from text prompts with Wan 2.5 Preview.
Developers can easily switch to production by using the Seedream 4.5 Playground API. Once you have tested your prompts in the UI, simply call the endpoint programmatically. You can find the API integration details and documentation on the Seedream 4.5 Edit API page.
Compared to its predecessor, Seedream 4.5 offers significantly better context understanding for long prompts and higher output fidelity. While Seedream 4.0 was limited to HD resolutions, Seedream 4.5 delivers true 4K texture details and handles complex instruction-based editing with greater accuracy.
To run the Seedream 4.5 Edit model, you must provide a prompt describing the desired changes and an image_url serving as the source. The model uses the source image as a structural reference while applying the semantic changes requested in your text prompt.
This specific page is for the Edit variant, which modifies existing images. If you need to generate images from scratch using the same high-quality architecture, you should use the Seedream 4.5 Text-to-Image model on RunComfy.
RunComfy hosts Seedream 4.5 on high-performance cloud GPUs to minimize latency. While 4K generations naturally take longer than 1K images due to the increased pixel count, the inference speed is optimized for interactive creative workflows and near-real-time iteration.
Seedream 4.5 is engineered for high structural consistency. It minimizes hallucinations by strictly adhering to the provided image_url for unedited areas, ensuring that only the elements specified in your prompt are altered while the rest of the composition remains intact.
RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.