Seedream 4.5
#43 in Text-to-ImagePicture Instruments · v4.5 · 8× · last seen Jun 30, 2026
2
Momentum
Seedream 4.5 is a proprietary text-to-image and image editing model by ByteDance (ByteDance Seed / BytePlus), released on December 3/4, 2025 as the successor to Seedream 4.0. The model generates native 2K/4K images from text prompts or reference images, supports multi-image editing with up to 10-14 reference images, and places particular emphasis on precise text rendering as well as character and object consistency across multiple images. It is primarily accessible via the official API (BytePlus ModelArk / Volcano Engine) as well as through third-party platforms such as Replicate, fal.ai, and WaveSpeedAI; standalone Fine-Tuning by end users is not officially documented.
Momentum trend
19.05.17.08.
Features
| Fine-Tuning | No official user-side fine-tuning documented; model is proprietary and closed-source (unlike open models such as Stable Diffusion) |
| Generation Time | Approx. 60 seconds for edits (fal.ai infrastructure); 4K about 50-100% slower than 2K |
| License | Proprietary (closed-source); commercial use of outputs typically allowed under platform terms of service |
| Max Resolution | Native 4K (up to 4096×4096, some sources report up to 4704×4704 pixels) |
| Platform | API via BytePlus ModelArk / Volcano Engine; also Replicate, fal.ai, WaveSpeedAI, etc. |
| Price | $0.04 per generated image (official API reference price, regardless of size) |
| Release Date | December 3-4, 2025 (model code seedream-4-5-251128) |