Octagram Creative

We ran two very quick tests on Weavy using the Nike Mercurial 10 Elite boot images.
Test 1: 2D to 3D:
- input: four images from different angles
- model used: Rodin 3D (the only one that allows for the above input)
- result: the 3D model is not usable for e-commerce purposes*

Test 2: Placing product in a contextual environment:
- input: one image and structured prompt
- models used: Gemini 2.5 Flash, Gemini 3 and Seedream v 4.5
- results: promising and can definitely be improved with additional prompt variations and refinements and parameters tweaking.

*𝘈𝘭𝘵𝘦𝘳𝘯𝘢𝘵𝘪𝘷𝘦 𝘢𝘱𝘱𝘳𝘰𝘢𝘤𝘩𝘦𝘴 𝘧𝘰𝘳 2D 𝘵𝘰 3D:
- 𝘨𝘦𝘯𝘦𝘳𝘢𝘵𝘪𝘯𝘨 𝘢 3𝘋 𝘱𝘳𝘰𝘥𝘶𝘤𝘵 𝘷𝘪𝘥𝘦𝘰 𝘧𝘳𝘰𝘮 2𝘋 𝘪𝘮𝘢𝘨𝘦𝘴 𝘸𝘩𝘪𝘤𝘩 𝘢𝘭𝘭𝘰𝘸𝘴 𝘧𝘰𝘳 𝘦𝘹𝘤𝘦𝘭𝘭𝘦𝘯𝘵 𝘱𝘳𝘰𝘥𝘶𝘤𝘵 𝘤𝘰𝘯𝘴𝘪𝘴𝘵𝘦𝘯𝘤𝘺, 𝘵𝘦𝘹𝘵𝘶𝘳𝘦 𝘢𝘯𝘥 𝘥𝘦𝘵𝘢𝘪𝘭𝘴 𝘱𝘳𝘦𝘴𝘦𝘳𝘷𝘢𝘵𝘪𝘰𝘯 (𝘢𝘴 𝘴𝘩𝘰𝘸𝘯 𝘩𝘦𝘳𝘦: https://www.facebook.com/share/v/1AGXpDumXF/)
- 𝘳𝘶𝘯𝘯𝘪𝘯𝘨 3𝘋 𝘱𝘳𝘰𝘥𝘶𝘤𝘵 𝘴𝘤𝘢𝘯 𝘢𝘯𝘥 𝘧𝘶𝘳𝘵𝘩𝘦𝘳 𝘳𝘦𝘧𝘪𝘯𝘪𝘯𝘨 𝘰𝘧 𝘵𝘩𝘦 3𝘋 𝘮𝘰𝘥𝘦𝘭

We are constantly exploring Generative AI use-cases grounded in real business needs. Product consistency, repeatable results and their applicability are key considerations in all our explorations as our aim is to transform experiments into scalable client-ready services.

We can now reliably produce 3D shoe visualisations using Generative AI alone from still images without any traditional 3D modelling (we never had this capability in house).
Our workflow can consistently generate 180° visualisations and we are confident that full 360° rotations will be possible soon.

For this particular test case we decided to use photos of a 𝐰𝐨𝐫𝐧 "Nike Manoa" boot (which belongs to one of our colleagues 😄) photographed by us in imperfect lighting conditions.
This way we can demonstrate how accurately the worn signs are reproduced and show the quality of the 3D visualisation derived from lower quality images.