ByteDance unveils faster Seedream 4.0 for image generation and editing
ByteDance's Seedream 4.0 brings text-to-image generation and image editing into a single system.
According to ByteDance, Seedream 4.0 is over ten times faster than Seedream 3.0. The model supports reference images, native control signals, and multi-image workflows. Users can generate images up to 4K resolution with flexible aspect ratios.

Text, tables, and diagrams are supposed to be clearer than before. Users can remove objects or adjust lighting directly with a text prompt. ByteDance reports that Seedream 4.0 takes the top spot in its internal "MagicBench" and "MagicArena" benchmarks for both text-to-image and single-image editing.

Developers and companies can use Seedream 4.0 through the Volcano Engine - Ark cloud platform, which was the first to roll out features like 4K output. The model is also built into ByteDance's Doubao chatbot and is available on no-code partner platforms including FAL.ai, Freepik, and Wavespeed.ai.
Pricing for Seedream 4.0 varies depending on platform and use case. On partner platforms, text-to-image or image editing costs about $0.03 per image, usually for images up to 1024×1024 pixels. Larger formats like 2K or 4K, as well as batch jobs, may cost more. The official Volcano Engine - Ark platform charges 0.2 yuan per image for Seedream 4.0 (about $0.028 at current rates). Pricing details for other regions and packages are available only after logging in or using the price calculator.
AI News Without the Hype – Curated by Humans
Subscribe to THE DECODER for ad-free reading, a weekly AI newsletter, our exclusive "AI Radar" frontier report six times a year, full archive access, and access to our comment section.
Subscribe nowRead on for the full picture.
Subscribe for hype-free coverage.
- Full access to every article on THE DECODER
- No ads
- Join the comments and community discussions
- A weekly AI news recap via mail
- 6x/year: "AI Radar" — deep dives on the AI topics that matter most
- Daily AI news, always up to date
- Our full ten-year archive
- Covered by a team with 10+ years in AI