Beginner struggling with Flux 2: SamplerCustomAdvanced node causing 15-minute generation times
PhilGrimnir · reddit · 2026-08-28
A new Flux 2 user reports a severe performance bottleneck where text-to-image generation takes 900 seconds (15 minutes), with 99% of the time stuck on the SamplerCustomAdvanced node. The setup uses an XFX RX 6750 XT (12GB VRAM) and 16GB RAM, with VAE encoding disabled for text-only prompts. Other models like Z-Turbo do not exhibit this delay. The user seeks explanations for the slowdown and tips for workflow optimization.
More from Multimodal
- Philipp Schmid Publishes Prompting Guide for Gemini Omni 1.1 Flash Video — _philschmid · 2026-08-28
- Fake digital-art speedpaint timelapses with MiniMax H3: workflow and full prompt shared — arthan1011 · 2026-08-28
- $5,000-Looking Ad Made with Just a Logo and a Prompt via MiniMax H3 — aziz4ai · 2026-08-28
- Google Launches Gemini Omni 1.1 Flash with 4K Upscaling and 10-Second Scene Extension — GoogleAI · 2026-08-28
- Runway Dev launches WAN 3.0 with multi-modal reference inputs — tlakomy · 2026-08-28
- Seedance 2.0 From $0.026/sec on Dreamina, Up to 76% Cheaper Than Rivals — azed_ai · 2026-08-28