How I generated 150+ yoga illustrations for my app as a solo founder with no design budget

How I generated 150+ AI illustrations as a solo founder with no design budget

When I started building yogakosh, I knew I needed a scalable visual system — not just a few images. 100+ pose illustrations. 150+ unique sequence covers. No designer. No budget for one.

I ended up building a pipeline across three tools:

• Recraft.ai for defining style and background removal
• DALL·E for generation and iterations
• Nano Banana pro to programmatically build prompts at scale based on yoga flow metadata (duration, style, focus area)

The part that took longest to figure out wasn't the tools — it was the PM decision to stop chasing perfection on edge cases. Complex poses with anatomical quirks were never going to be solved cleanly by AI. So I prioritized high-confidence poses and let audio cues carry the nuance for the more complex pose variations. Faster to ship, better user experience.

Went from 15–30 min per image manually to batch generation + review.

Full breakdown with workflows and prompt evolution here: https://www.yogakosh.com/blog/from-1-to-100-automating-asset-creation-for-my-yoga-app

Happy to answer questions if anyone's dealing with asset creation at scale as a solo builder.