ByteDanceByteDance·🖌️ Inpainting·New

Seedream V5.5

anonymized
Try on Venice.ai ↗
Quick reference
Seedream V5.5 — TLDR
  • 🏢 ByteDance Seed's image editing model, released November 2026.
  • 💬 Natural-language local edits; unedited regions stay intact pixel-level.
  • 🔧 Accepts up to 10 input images as editing references.^catalog^
  • 📏 Output tiers at 1K, 1.5K, and 2K.^catalog^
  • 🆕 Stable multi-turn editing across many successive rounds.
  • 🎯 ByteDance reports +55.62% GSB realism gain over prior version.
  • 📚 Improved rendering of dense, small and multilingual text.
💰 Pricing
$0.025
per edit
📅 On Venice since
Nov 10, 2026
6 days ago
Provider

ByteDance is a Chinese internet technology company headquartered in Beijing's Haidian district, with an associated variable-interest entity incorporated in the Cayman Islands. Best known as a consumer internet operator at enormous scale, it has also become a…

Read full profile →
36 models on Venice
28 video · 4 image · 4 inpaint
Since Nov 5, 2025

About this model

Seedream V5.5 Edit is the image-editing variant of ByteDance Seed's Seedream 5.5 family, released in November 2026 alongside its text-to-image counterpart Seedream V5.5. It takes an existing image plus a natural-language instruction, and per the Venice catalog description can be guided by up to ten reference images, producing results at 1K, 1.5K, or 2K.

The headline change over earlier Seedream editors is pixel-level stability during multi-turn work: ByteDance describes targeted local changes that leave unedited regions intact, holding detail and color across long chains of successive edits. That directly addresses the drift problem in iterative editing, where later instructions degrade earlier ones.

Compared with its same-family predecessor Seedream V4.5 and the Seedream V5 Pro editor, ByteDance reports a substantial overall improvement, with expert GSB (Good/Same/Bad) pairwise scoring positive against the previous version in all four text-structure and realism comparisons, reaching +55.62% for realism on image-to-image tasks. The company also reports better generation and layout of densely arranged small text, including complex character shapes.

Within the catalog, Seedream 5.5 covers both generation and editing entry points, and sits beside ByteDance's video line including Seedance 2.5 Edit. The quality figures above are ByteDance's own vendor-reported evaluations rather than independent measurements.

This About section is AI-generated from public sources (Claude Opus 5), with no human editing. It may contain inaccuracies — verify critical details against the sources listed above.

Data sources: Venice API · HuggingFace · Wikipedia — enrichment updated 6d ago