Text this: Visual tracking by matching points using diffusion model