TikTok’s parent company, ByteDance, has reportedly launched its new image generation tool, Seedream 4.0. The company claims that it outperforms Google DeepMind’s popular Gemini 2.5 Flash Image (also known as “Nano Banana”). The company’s Seed department stated that Seedream 4.0 boasts strong image editing capabilities, positioning it in direct competition with Google’s tool, which has garnered widespread acclaim for its consistent and accurate image editing. ByteDance claims that Seedream 4.0 beat Gemini 2.5 Flash Image on its internal evaluation benchmark, MagicBench, with better performance in prompt adherence, alignment, and aesthetics. However, the company has not published these results in an official technical report. This comes after Google’s Nano Banana AI 3D figurine trend exploded on social media. This image-generating tool enables anyone to create customisable, hyper-realistic mini figurines with ease. Their playful design, creative flexibility, and shareable appeal have made them a favourite among creators and collectors. ByteDance said Seedream 4.0 introduces a new architecture that accelerates raw image inference by more than 10 times compared with earlier models, making it faster to use. Early user feedback has highlighted the editing tool’s precision, with quick image adjustments possible through text prompts.The tool is available domestically via ByteDance’s Jimeng and Doubao AI apps, and to enterprise customers through its Volcano Engine cloud platform.
How Bytedance’s Seedream compares to Nano Banana
According to a report by the South China Morning Post, AI benchmark company Artificial Analysis has noted that Seedream 4.0 marked a “significant evolution” from earlier ByteDance tools by combining the text-to-image functions of Seedream 3.0 with the editing features of SeedEdit 3.0, while maintaining the exact price of $30 per 1,000 generations.At present, Gemini 2.5 Flash Image leads the company’s rankings for both text-to-image generation and image editing, while Seedream 4.0 has not yet been assessed. Seedream 3.0 currently stands fifth in text-to-image and sixth in editing.On Fal.ai, a global generative media hosting site, Seedream 4.0 costs US$0.03 per generated image, compared with US$0.039 for Gemini 2.5 Flash Image.China has rapidly adopted AI-driven image and video generation, with competition heating up among tech companies, including Kuaishou Technology and Tencent Holdings. The government has supported the sector by recognising copyright for AI-generated content in late 2023, while also mandating labels on such material.The market has also seen an influx of competitively priced products, such as Vidu, a video generation tool created by Tsinghua University and start-up Shengshu AI. Earlier this week, Vidu added a reference-to-image feature internationally, enabling generation from up to seven reference images at 0.62 yuan (US$0.09) per image. In comparison, Gemini 2.5 Flash Image supports up to nine references.