Can you save on LLM tokens using images instead of text?

What happens if you convert your prompts to an image before running?