Should style and speed LoRAs go into the latent upscale pass in two-pass workflows?

joseph_jojo_shabadoo · reddit · 2026-09-02

A Reddit user asks two practical questions about two-pass latent scaling workflows with H3: are style/character LoRAs supposed to be piped into the latent upscale pass too, or just the native pass? And should/could a speed-up LoRA also be piped into the upscale pass?

Original post →

More from Multimodal

Multimodal channel →