A ComfyUI based model: https://github.com/fofr/cog-comfyui-image-merge
You can merge two images and use an optional third as a controlnet input.
Merge options include: full – each image applies to the whole image left_right - first on left, second on right top_bottom – first on top, second on bottom
When using controlnet, openpose, depth and canny nets are applied.
The 2x upscaler uses controlnet tile to upscale.