jeu. 30 juil. 2026
Deploymates
image generation

Workflows ComfyUI pour la génération d'images produit

Construisez des pipelines d'images reproductibles pour landing pages SaaS sans abonnement Midjourney.

TL;DR

ComfyUI avec SDXL ou Flux donne aux indie hackers un contrôle total sur mockups produit et assets marketing à ~0,02 $/image sur un VPS GPU à 20 $/mois — 10× moins cher que les outils par abonnement pour les équipes générant 500+ images/mois.

Sam ChenSam Chen10 min de lecture
Workflows ComfyUI pour la génération d'images produit

ComfyUI changed how indie hackers produce marketing assets. Instead of $30/mo Midjourney subscriptions or stock photo hunts, you build reproducible node graphs that generate on-brand product mockups, hero images, and social creatives at pennies per render.

Why ComfyUI over Midjourney for SaaS marketing?

ComfyUI vs subscription tools for indie teams
FactorComfyUI + SDXL/FluxMidjourney
Cost at 500 images/mo~$10 GPU$30–60 subscription
ReproducibilityFull workflow JSONPrompt-only, inconsistent
Brand controlLoRA, ControlNetLimited
Batch processingNativeManual
Learning curveSteep (2–3 days)Low

ComfyUI workflow for product hero images

  1. Load SDXL or Flux checkpoint
  2. Add ControlNet with your UI screenshot for layout guidance
  3. Apply LoRA trained on your brand palette (optional, 20 images enough)
  4. Set resolution 1024×1024, batch size 4 for variants
  5. Upscale best output with Real-ESRGAN 4×
  6. Export WebP at 85% quality for landing pages

Example: AI-generated SaaS hero aesthetic (designer workflow)

View original →

GPU options for ComfyUI in 2026

  • RunPod RTX 4090: ~$0.44/hr — best for batch sessions
  • Local M-series Mac: 16GB+ unified memory, slower but free
  • Vast.ai spot instances: cheapest for overnight batch jobs
  • Hetzner GPU (limited availability): EU data residency

ComfyUI beginner workflow — product mockups

View original →

Frequently asked questions

Flux or SDXL for SaaS landing page images?
Flux for photorealism and text-in-image. SDXL for speed on consumer GPUs. Most indie teams use Flux at 1024px for heroes, SDXL for social thumbnails.
Can designers without coding skills use ComfyUI?
Yes — the node UI is visual. Start with community workflows on OpenArt. Coding helps for automation and API integration, not basic generation.
How do I keep generated assets on-brand?
Train a LoRA on 15–30 brand screenshots. Use ControlNet with wireframes. Store workflow JSON in git for reproducibility.

Articles similaires