Bytedance Seedream V4.5 API

Edit
Text To Image

Endpoint: POST https://fal.run/fal-ai/bytedance/seedream/v4.5/edit Endpoint ID: fal-ai/bytedance/seedream/v4.5/edit

Try it in the Playground

Run this model interactively with your own prompts.

Quick Start

import fal_client

def on_queue_update(update):
    if isinstance(update, fal_client.InProgress):
        for log in update.logs:
           print(log["message"])

result = fal_client.subscribe(
    "fal-ai/bytedance/seedream/v4.5/edit",
    arguments={
        "prompt": "Replace the product in Figure 1 with that in Figure 2. For the title copy the text in Figure 3 to the top of the screen, the title should have a clear contrast with the background but not be overly eye-catching.",
        "image_urls": [
            "https://storage.googleapis.com/falserverless/example_inputs/seedreamv45/seedream_v45_edit_input_1.png",
            "https://storage.googleapis.com/falserverless/example_inputs/seedreamv45/seedream_v45_edit_input_2.png",
            "https://storage.googleapis.com/falserverless/example_inputs/seedreamv45/seedream_v45_edit_input_3.png"
        ]
    },
    with_logs=True,
    on_queue_update=on_queue_update,
)
print(result)

Input Schema

prompt

string

required

The text prompt used to edit the image

image_size

ImageSize | Enum

default:"[object Object]"

The size of the generated image. Width and height must be between 1920 and 4096, or total number of pixels must be between 25601440 and 40964096.Possible values: square_hd, square, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9, auto_2K, auto_4K

num_images

integer

default:"1"

Number of separate model generations to be run with the prompt. Default value: 1Range: 1 to 6

max_images

integer

default:"1"

If set to a number greater than one, enables multi-image generation. The model will potentially return up to max_images images every generation, and in total, num_images generations will be carried out. In total, the number of images generated will be between num_images and max_images*num_images. The total number of images (image inputs + image outputs) must not exceed 15 Default value: 1Range: 1 to 6

seed

integer

Random seed to control the stochasticity of image generation.

sync_mode

boolean

default:"false"

If True, the media will be returned as a data URI and the output data won’t be available in the request history.

enable_safety_checker

boolean

default:"true"

If set to true, the safety checker will be enabled. Default value: true

image_urls

list<string>

required

List of URLs of input images for editing. Presently, up to 10 image inputs are allowed. If over 10 images are sent, only the last 10 will be used.

Output Schema

images

list<Image>

required

Generated images

Input Example

{
  "prompt": "Replace the product in Figure 1 with that in Figure 2. For the title copy the text in Figure 3 to the top of the screen, the title should have a clear contrast with the background but not be overly eye-catching.",
  "image_size": "auto_4K",
  "num_images": 1,
  "max_images": 1,
  "sync_mode": false,
  "enable_safety_checker": true,
  "image_urls": [
    "https://storage.googleapis.com/falserverless/example_inputs/seedreamv45/seedream_v45_edit_input_1.png",
    "https://storage.googleapis.com/falserverless/example_inputs/seedreamv45/seedream_v45_edit_input_2.png",
    "https://storage.googleapis.com/falserverless/example_inputs/seedreamv45/seedream_v45_edit_input_3.png"
  ]
}

Output Example

{
  "images": [
    {
      "url": "https://storage.googleapis.com/falserverless/example_outputs/seedreamv45/seedream_v45_edit_output.png"
    }
  ]
}

Endpoint: POST https://fal.run/fal-ai/bytedance/seedream/v4.5/text-to-image Endpoint ID: fal-ai/bytedance/seedream/v4.5/text-to-image

Try it in the Playground

Run this model interactively with your own prompts.

Quick Start

import fal_client

def on_queue_update(update):
    if isinstance(update, fal_client.InProgress):
        for log in update.logs:
           print(log["message"])

result = fal_client.subscribe(
    "fal-ai/bytedance/seedream/v4.5/text-to-image",
    arguments={
        "prompt": "A selfie of a cat, with the cat as the protagonist. The setting is twilight at the Eiffel Tower. The cat is happy, holding a piece of baklava in its paw. The photo has a slight motion blur and is slightly overexposed. From a selfie angle, with a bit of motion blur, the overall image presents a sense of calm madness. The text \"Seedream 4.5 is on fal\" should be written on the picture at the top in clearly visible font and crisp lettering. The image has a 4:3 aspect ratio"
    },
    with_logs=True,
    on_queue_update=on_queue_update,
)
print(result)

Input Schema

prompt

string

required

The text prompt used to generate the image

image_size

ImageSize | Enum

default:"[object Object]"

num_images

integer

default:"1"

Number of separate model generations to be run with the prompt. Default value: 1Range: 1 to 6

max_images

integer

default:"1"

seed

integer

Random seed to control the stochasticity of image generation.

sync_mode

boolean

default:"false"

If True, the media will be returned as a data URI and the output data won’t be available in the request history.

enable_safety_checker

boolean

default:"true"

If set to true, the safety checker will be enabled. Default value: true

Output Schema

images

list<Image>

required

Generated images

seed

integer

required

Seed used for generation

Input Example

{
  "prompt": "A selfie of a cat, with the cat as the protagonist. The setting is twilight at the Eiffel Tower. The cat is happy, holding a piece of baklava in its paw. The photo has a slight motion blur and is slightly overexposed. From a selfie angle, with a bit of motion blur, the overall image presents a sense of calm madness. The text \"Seedream 4.5 is on fal\" should be written on the picture at the top in clearly visible font and crisp lettering. The image has a 4:3 aspect ratio",
  "image_size": "auto_2K",
  "num_images": 1,
  "max_images": 1,
  "sync_mode": false,
  "enable_safety_checker": true
}

Output Example

{
  "images": [
    {
      "url": "https://storage.googleapis.com/falserverless/example_outputs/seedreamv45/seedream_v45_t2i_output.png"
    }
  ],
  "seed": 42
}

ByteDance’s Seedream 4.5 transforms existing images through natural language instructions at $0.04 per edit, processing up to 10 reference images simultaneously for complex multi-source compositions. Trading simple single-image workflows for sophisticated context-aware editing, this unified architecture references multiple sources, copies specific elements between images, and maintains spatial relationships without manual masking. Built for e-commerce teams assembling product composites, designers prototyping layout variations, and marketing workflows requiring consistent brand element integration across visuals. Built for: Multi-image product composites | Layout prototyping with text overlays | Brand asset integration workflows

Natural Language Editing Without Layers

Seedream 4.5 consolidates image generation and editing into a single architecture that interprets spatial references directly from your prompt. Instead of requiring layer masks or selection tools, you describe edits using natural language - “replace the product in Figure 1 with that in Figure 2” or “copy the text from Figure 3 to the top with clear contrast.” What this means for you:

Multi-source composition: Reference up to 10 images per edit, enabling complex workflows like product swaps, text overlay copying, and element positioning across multiple source files
Context-aware transformations: The model maintains depth, perspective, and lighting consistency when integrating elements from different sources - no manual blending required
Resolution flexibility: Output up to 4 megapixels (2048x2048 maximum) with configurable dimensions between 1920px and 4096px on either axis
Batch generation control: Run 1-6 separate generations per request, with optional multi-image output (up to 6 images per generation) for exploring variations

Performance That Scales

Seedream 4.5 processes edits in approximately 60 seconds on fal infrastructure, with pricing structured for production workflows requiring multiple reference images.

Metric	Result	Context
Inference Speed	~60 seconds	Standard processing time per edit on fal
Cost per Edit	$0.04	25 edits per $1.00 on fal
Max Reference Images	10 images	Multi-source composition capability (last 10 used if more provided)
Max Resolution	4MP (2048x2048)	Configurable dimensions between 1920-4096px per axis

Technical Specifications

Spec	Details
Architecture	Seedream 4.5
Input Formats	Image URLs (up to 10), text prompt
Output Formats	PNG images via URL or data URI
Resolution Range	1920-4096px per axis, 2560×1440 to 4096×4096 total pixels
License	Commercial use via fal Partner agreement

API Documentation

How It Stacks Up

Bytedance Seedream v4 Edit - Seedream 4.5 expands multi-image input capacity from v4’s baseline while maintaining the unified editing architecture. Both versions handle natural language spatial instructions, with v4.5 prioritizing higher reference image limits for complex composition workflows. Bytedance Seededit v3 - Seedream 4.5 consolidates generation and editing into a single model architecture, trading v3’s specialized editing focus for broader capability coverage. Seededit v3 remains purpose-built for pure image-to-image transformation workflows without generation requirements. NAFNet-deblur - Seedream 4.5 handles multi-image composition and semantic editing through natural language, making it ideal for layout assembly and element integration. NAFNet-deblur specializes in single-image restoration tasks like blur removal and artifact correction where semantic understanding isn’t required.

Bytedance — Image Generation

Limitations

image_size restricted to: square_hd, square, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9, auto_2K, auto_4K
num_images range: 1 to 6
max_images range: 1 to 6
Content moderation via safety checker

Video Generation

Image Generation

Audio

Vision

3D

Bytedance Seedream V4.5 API

Try it in the Playground

Quick Start

Input Schema

Output Schema

Input Example

Output Example

Try it in the Playground

Quick Start

Input Schema

Output Schema

Input Example

Output Example

Natural Language Editing Without Layers

Performance That Scales

Technical Specifications

How It Stacks Up

Limitations

Video Generation

Image Generation

Audio

Vision

3D

Try it in the Playground

​Quick Start

​Input Schema

​Output Schema

​Input Example

​Output Example

Try it in the Playground

​Quick Start

​Input Schema

​Output Schema

​Input Example

​Output Example

​Natural Language Editing Without Layers

​Performance That Scales

​Technical Specifications

​How It Stacks Up

​Related

​Limitations

Quick Start

Input Schema

Output Schema

Input Example

Output Example

Quick Start

Input Schema

Output Schema

Input Example

Output Example

Natural Language Editing Without Layers

Performance That Scales

Technical Specifications

How It Stacks Up

Related

Limitations