ComfyUI just dropped a Lora training node. If we can actually train LoRAs directly inside ComfyUI, that’d save so much hassle—back in the day you had to keep switching tools and environments between training and inference, super fragmented. Running the whole pipeline from training to generating images in one interface would really lower the barrier for newbies. Mainly just wondering if anyone’s already tried this training node and if it’s actually any good in practice.
Yeah, going through the whole training-to-inference pipeline is honestly so satisfying.
My only worry is that if the node encapsulation is too rigid, it’ll be a pain to tweak the parameters.
Gonna wait for some hands-on reviews before deciding if it’s worth the hassle.
ComfyUI ecosystem is seriously competitive right now.
Training and inference in one complete pipeline feels great, but I’m worried the nodes are too tightly packed to tweak the parameters easily.
1 Like