Training Krea2 character LoRAs on a 4060 8GB in ~40 minutes
BigBullshitta · reddit · 2026-09-22
A Reddit user demonstrated that a consumer RTX 4060 8GB with 32GB system RAM is enough to train character LoRAs for Krea2 locally. Using One Trainer with the Krea 2 Raw base converted to int8 w8A8, a 35-image dataset, 50 epochs at lr 0.0001, the whole run took 1h1m including the base model download (40 minutes for the LoRA itself).
The same machine runs Krea2 Turbo with a community 4-step LoRA, generating 1.4MP images in 17 seconds — versus 15-20s per SDXL image two years ago. Local image generation barriers have dropped dramatically.
More from Infra
- A Wild Async RL Config: 30 Steps x 25K Rollouts Per Step at Parallelism 4 — willcb · 2026-09-22
- AMD's market cap went from $2B to $1T in the 11 years since Lisa Su became CEO — xiaosun86 · 2026-09-22
- Blackstone poured ~$100B into data centers this year, telling investors this is not the dot-com bubble — JOBhakdi · 2026-09-22
- Modular Releases Free LLM Inference Handbook With 20+ Interactive Visualizations — carrycooldude · 2026-09-22
- Naveen Rao at All-In Summit: AI is extremely inefficient and every computer since 1945 is built wrong — NaveenGRao · 2026-09-22
- GKE Pod snapshots cut AI inference cold starts by 89%, loading 70B models in 37s — rseroter · 2026-09-22