do you think it would be possible to modify this model to generate images?

[Manni1000]

i think with some small chages it could make images. and could replicate something like the new google gemini image/ open aigpt4o image. what do you think about that?

[wzk1015]

Good idea! Actually we have explored extending the Mono-InternVL structure to image generation and proposed SynerGen-VL (https://huggingface.co/papers/2412.09604). It is also accepted to CVPR 2025 and will be open-sourced soon. We are still continuously working on more advanced and efficient monolithic MLLMs. Stay tuned!