Structuring ComfyUI for Batch Product Shots Without Drift
ayanbhadwa · reddit · 2026-09-01
The author discusses a ComfyUI workflow for controlled product shot variation to avoid misleading customers. Key points include:
- Locking Core Features: Use masks and reference conditioning to fix silhouette, materials, and logos.
- Decoupling Variables: Expose environment, lighting, and camera angle as separate variables for incremental testing.
- Metadata Tracking: Save model, prompt, and variables for traceability.
The author uses Atlas Cloud for model routing and debates whether to enforce product fidelity pre-generation (hard constraints) or post-generation (QC), balancing visual novelty with accuracy to avoid setting wrong expectations.
More from Multimodal
- Any lora for characters? — NefariousnessFun4043 · 2026-09-01
- Kijai released FastH3 checkpoint, but standard ComfyUI workflows won't run it yet — bub000 · 2026-09-01
- OpenAI Sol is the first model that truly understands music, tester says — evilsocket · 2026-09-01
- Three months running Latentsync on 8GB locally: the real cost of "free" — cloudybrain07 · 2026-09-01
- Generating Cinematic Titles with "Reverse Calligraphy" Concept — umesh_ai · 2026-09-01
- Handwritten Code vs. AI-Generated Code in Generative Art — adamho · 2026-09-01