Ideogram V4 LoRA Trainer

New

Train custom image styles, characters, or brand visuals and generate consistent assets with Ideogram V4.

A short name shown later in Ideogram 4.0.

Optional phrase to reuse in Ideogram prompts.

Used for images without caption files. Include a unique phrase here if you want a trigger word in prompts later.

1000

More steps can capture the dataset more strongly, but cost and training time scale linearly.

Auto chooses the largest common no-upscale crop. Presets crop images without upscaling.

Balance: 0 tokens

Training Dataset ZIP is required

Your trained LoRA weights and usage instructions will appear here