Online image training

FLUX LoRA trainer online

Train a portable image LoRA without a Google Colab notebook, Python environment, CUDA setup, or local GPU. NiceVois turns a known training workflow into a small upload form and gives you the resulting .safetensors file.

What you provide

• 8–40 clear images you own or have permission to use — JPG, PNG, WEBP, HEIC and other common formats all work.
• A short name for the subject.
• Whether the images represent a person, product, character, or visual style.

What NiceVois handles

NiceVois validates the images, writes a consistent caption set, builds the training configuration, sends the job to a separate image-training GPU, checks the returned package, and makes the portable result downloadable. Training runs on the Apache-2.0 Flex.1-alpha base model.

What you receive

The ZIP contains the trained .safetensors LoRA, the exact manifest used for the run, a short usage guide, and any generated validation sample. The file is not locked inside a NiceVois account.

Prepare the images before training

Good training data matters more than adding near-duplicate pictures. Use varied angles, lighting, distance, and background while keeping the intended subject consistent. Read the image LoRA dataset guide before your first run.