Full-AI pipeline turns Blender city into stylized video via GPT, DepthAnything and Hailuo
Hailuo_AI · x · 2026-09-09
A Japanese developer (kiyoshishin) demonstrated a fully AI-driven video creation pipeline with promising results:
- Built a steampunk city with BlenderMCP + GPT (Astra/Codex), then converted it to a line-art style
- Pipeline: extract 3 keyframes and restyle them into concept art with GPT Image; run the 15s video through ComfyUI + DepthAnything for depth-based stylization; feed the results into Hailuo (MiniMax) with prompts drafted by Astra
- Lessons learned: feeding the raw video plus images yields the 3D output unchanged; with 6 reference images scenes drift subtly, so 3 images plus AI freedom works best
- Everything, including the making-of documentation, was done inside Codex (Astra)
Related event: Developers Build Full-AI Pipeline for Steampunk City Videos(2 posts)→
More from coding & agent
- ComfyMax v0.2: a simpler frontend for local ComfyUI + MiniMax H3 workflow — DanielVeres · 2026-09-09
- AI connectors and browser use let perfectionists delegate without actually delegating — tushaarmehtaa · 2026-09-09
- Hackathon-winning memory system rebuilt as zero-dependency MCP server with SQLite knowledge graph — BoringCelebration405 · 2026-09-09
- Veteran dev claims a solo GTA-scale game in under 6 months is possible with UE plus AI tools — D3VAUX · 2026-09-09
- One good model beats a thousand sub-agents: a sharp take on agent architecture — tekbog · 2026-09-09
- Replacing a Telegram-to-Second-Brain summarizer workflow with Claude Code — sebpaquet · 2026-09-09