Skip to main content
seedream-4.5 is a refinement release over seedream-4. Improvements concentrate on prompt adherence, edge fidelity (less “AI cleanup” on hand-drawn elements), and Chinese in-image text consistency. Per-image cost is slightly higher than seedream-4. The right pick when seedream-4 was already working and you want the increment, not when you’re choosing among tiers for the first time. Pricing: $0.030 / image — see the rate card. Failures don’t bill.

​
Protocols

ProtocolPath
OpenAI ImagesPOST https://llm.bytespike.ai/v1/images/generations

​
Quickstart

curl https://llm.bytespike.ai/v1/images/generations \
  -H "Authorization: Bearer $BYTESPIKE_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "seedream-4.5",
    "prompt": "国潮风插画，敦煌色彩，飞天主角居中，背景为佛窟纹样",
    "n": 1,
    "size": "1024x1024"
  }'
Python and TypeScript invocations match the pattern shown on seedream-4 — just swap the model field.

​
Capabilities

CapabilitySupported
Text-to-image✅
Image-to-image (with source_image)✅
n ≥ 2 batch generation✅
size 1024×1024 / 1024×1792 / 1792×1024✅
Chinese in-image typography✅ stronger than 4
response_format url / b64_json✅
Modalityimage
Capability bucketimage_generate

​
When to use

  • You’re already on seedream-4 and want the next iteration’s improvements.
  • Tight prompt adherence required — campaign work where the prompt specifies layout details that can’t drift.
  • Chinese in-image text — small per-image quality gain shows up most on text-heavy posters.
When not to use:
  • You’re starting from scratch and not sure between tiers — pick seedream-4 first, upgrade only if you hit a real gap.
  • Cost-sensitive volume — drop to seedream-v5lite.

​
Next