Fake 3D-print figurine photos with one long prompt: author shares full Krea2 + MiniMax H3 workflow
alisitskii · reddit · 2026-10-05
A Reddit user showcased a convincing "fake 3D print" AI pipeline: Krea2 generates a photorealistic indoor still life of a Lara Croft figurine, with the prompt explicitly describing speckled print-like material, visible layer lines, shallow depth of field and natural window light; MiniMax H3 then converts it into handheld-smartphone-style video.
The full prompts are shared. Key tricks:
- Krea2: lean hard into tactile print artifacts and mundane domestic framing to sell realism
- MiniMax H3: describe irregular hand tremor, autofocus breathing, exposure drift and room-tone audio instead of cinematic camera moves, and set non-diegetic music to N/A
The recipe is directly reusable for product mockups, figurine concepts or mimicking UGC video texture.
More from Multimodal
- Rebel Filmmaker Damon Packard Goes All-In on AI Video, and Nobody's Watching — Kyrannio · 2026-10-05
- BAAI's AREX-2, an agent-focused multimodal model on Qwen3.5, trends on Hugging Face — BAAI · 2026-10-05
- Reinventing realtime datamoshing with shaders, DCT residuals and feedback loops — pixlpa · 2026-10-05
- Open-source refreshing_extender workflow breathes natural motion into static video generation — roychodraws · 2026-10-05
- ElevenLabs v4 + Dreamina Called a Perfect Pairing for Virtual Influencer Content — AIandDesign · 2026-10-05
- AI-generated virtual podcasters return with new episode on Taylor Swift and Aussie snacks — AIandDesign · 2026-10-05