AI Art Generation Pipeline: 100+ Daily Images on Autopilot
End-to-end n8n pipeline for generating cyberpunk AI art at scale — GPT-4 prompts, Midjourney image generation via async webhook, Recraft upscaling, all managed through Airtable state machines.
30x
Faster than manual
90%+
Labor cost reduction
4
Variations per prompt
$0.10–0.30
Cost per full cycle
The Problem
- Content creators needed hundreds of unique, high-quality cyberpunk images at scale.
- Manual prompt engineering took 5 minutes per prompt.
- Managing multiple API services, tracking generation status, and organizing outputs required constant attention.
- Human fatigue created declining prompt quality and repetitive patterns across extended sessions.
The Solution
Three-stage pipeline.
Stage 1 — Prompt Generation
- GPT-4o-mini receives a 1,500+ word instruction set with 20 setting options, 20 character types, and 20+ visual elements.
- Randomly combines elements to ensure uniqueness.
- A second GPT-4 agent converts the static Midjourney prompt into a dynamic Sora video sequence description.
- Both stored in Airtable.
Stage 2 — Image Generation
- Airtable trigger fires on new prompts, sends to Kie.ai Midjourney API with async callback webhook, generates 4 variations per prompt.
- Webhook receives completion, JavaScript parses 4 image URLs, each stored as separate Airtable record.
Stage 3 — Selective Upscaling
- Human quality gate marks images for upscaling.
- Recraft crisp upscale API processes via async callback.
- Final images stored in Upscaled Images table.
Results
Prompt creation: 5 minutes manual → 10 seconds automated — 30x faster
100-prompt batch: 8+ hours manual → 15 minutes automated
Millions of unique element combinations prevent repetitive output patterns
Async webhook architecture allows parallel generation without blocking
Human quality gate between generation and upscaling ensures only best images receive processing
Tech Stack
Want something like this?
Tell me what you're building. I'll tell you honestly if I can help.