Seedream Image

Seedream Image

Generate AI images using ByteDance Seedream 4.5. Use when you need to: (1) create images from text descriptions, (2) transform images with style transfer, or (3) generate high-quality artistic or realistic images.

Input

Provide input as JSON:

jsonCopy code
{
  "prompt": "Text description of the image you want to generate or edit",
  "image_size": "Image size: square_hd, square, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9, auto_2K, auto_4K (default: auto_2K)",
  "num_images": "Number of images to generate (1-10, default: 1)",
  "reference_image": "<file-id>",
  "enable_safety_checker": "Enable safety checker (true/false, default: true)"
}

Note on File Input:

• reference_image is optional and requires a file ID (format: df-xxxxx) if provided
• How to get file ID:
  1. Upload your reference image to Refly using refly file upload <file-path>
  2. Copy the returned file ID from the upload response
  3. Use this file ID in the input JSON
• Omit reference_image for pure text-to-image generation
• Provide reference_image for image-to-image editing (supports up to 10 images)

API Request Structure

This skill uses Seedream 4.5 API with the following structure:

Text-to-Image

jsonCopy code
{
  "prompt": "Your prompt here",
  "image_size": "auto_2K",
  "num_images": 1,
  "max_images": 1,
  "enable_safety_checker": true
}

Image-to-Image (With Reference)

jsonCopy code
{
  "prompt": "Edit instructions here",
  "image_size": "auto_4K",
  "num_images": 1,
  "max_images": 1,
  "enable_safety_checker": true,
  "image_urls": [
    "URL_of_reference_image_1",
    "URL_of_reference_image_2"
  ]
}

Notes:

• When image_urls is empty/not provided → Text-to-Image mode
• When image_urls is provided → Image-to-Image mode
• Supports up to 10 input images
• Total images (input + output) must not exceed 15
• image_size options: square_hd, square, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9, auto_2K, auto_4K

Execution (Pattern A: File Generation)

Step 0 (Optional): Upload Reference Image

bashCopy code
# Upload a reference image for image-to-image generation
REF_RESULT=$(refly file upload /path/to/reference.jpg)
REF_FILE_ID=$(echo "$REF_RESULT" | jq -r '.payload.fileId')
echo "Reference image file ID: $REF_FILE_ID"

Step 1: Run the Skill and Get Run ID

Example 1: Text-to-Image

bashCopy code
RESULT=$(refly skill run --id skpi-m56jo9am75xgijga62hd7t8g --input '{
  "prompt": "A beautiful mountain landscape at sunset with vibrant colors",
  "image_size": "landscape_16_9",
  "num_images": 1,
  "enable_safety_checker": true
}')
RUN_ID=$(echo "$RESULT" | jq -r '.payload.workflowExecutions[0].id')

Example 2: Image-to-Image Editing

bashCopy code
# Use the REF_FILE_ID from Step 0
RESULT=$(refly skill run --id skpi-m56jo9am75xgijga62hd7t8g --input '{
  "prompt": "Transform this into anime style with vibrant colors and dramatic lighting",
  "reference_image": "'"$REF_FILE_ID"'",
  "image_size": "auto_4K",
  "num_images": 1,
  "enable_safety_checker": true
}')
RUN_ID=$(echo "$RESULT" | jq -r '.payload.workflowExecutions[0].id')

Step 2: Open Workflow in Browser and Wait for Completion

bashCopy code
open "https://refly.ai/workflow/c-i51ssnvhcjdou7l0h3ote7d9"
refly workflow status "$RUN_ID" --watch --interval 30000

Step 3: Download and Show Result

bashCopy code
# Get files from this run
FILES=$(refly workflow toolcalls "$RUN_ID" --files --latest | jq -r '.payload.files[]')

# Download and open each file
echo "$FILES" | jq -c '.' | while read -r file; do
  FILE_ID=$(echo "$file" | jq -r '.fileId')
  FILE_NAME=$(echo "$file" | jq -r '.name')
  if [ -n "$FILE_ID" ] && [ "$FILE_ID" != "null" ]; then
    refly file download "$FILE_ID" -o "$HOME/Desktop/${FILE_NAME}"
    open "$HOME/Desktop/${FILE_NAME}"
  fi
done

Expected Output

• Type: Image
• Format: .png/.jpg image file
• Location: ~/Desktop/
• Action: Opens automatically to show user

Rules

Follow base skill workflow: ~/.claude/skills/refly/SKILL.md