ByteDance Seedream 5.5 hands-on: survives 100 rounds of consecutive edits, costs $0.04 per 2K edit
量子位 · wechat · 2026-11-10
QbitAI's hands-on review of ByteDance's Seedream 5.5 (live in Doubao, Jimeng, Doubao Work) versus GPT-Image 2.6:
- 100-round consecutive editing: in a stress test incrementing a watch's time each round, GPT-Image 2.6 started distorting around round 10 and degraded badly; Seedream 5.5 stayed far more faithful.
- Multi-image fusion & transparent PNGs: merged three shots with mismatched lighting into a coherent ad visual; composited five semi-transparent assets with correct translucency, refraction and colored shadows.
- Realism & tiny text: cinematic frames with natural warm/cool light transitions and shallow depth of field; dense multilingual science posters rendered with virtually no garbled text; action transfer and multi-person pose consistency also work.
- Layer splitting: can decompose a finished ad into editable layers, swapping market-specific copy while freezing packaging and composition — useful for global brand localization. GIF generation supported.
- Pricing: keeps the input+output billing structure; a 2K square edit costs about 0.28 RMB ($0.04), less than half of Seedream 5.0 Pro.
More from Multimodal
- Hailuo showcases H3Image's vintage camera look—with hilariously spooky results — Hailuo_AI · 2026-11-11
- Microsoft launches MAI-Transcribe-2, claiming 10x faster transcription at $0.10 per hour — mustafasuleyman · 2026-11-11
- Hailuo spotlights H3Image demos recreating nostalgic 90s anime aesthetic — Hailuo_AI · 2026-11-11
- Carbon demo: hand-written CUDA path tracer plus MuJoCo physics powers 60-second one-take — fofrAI · 2026-11-11
- Runway explores real-time responsive video for learning, creating and collaborating — runwayml · 2026-11-11
- MiniMax H3-Image merges 16 reference images into one cinematic scene in 10 seconds — Hailuo_AI · 2026-11-11