Spaces:

---

bytedance-research

/

Lance

Like 114

Sleeping

using gradio server with a custom frontend see docs: https://huggingface.co/blog/introducing-gradio-server

@ffy2000

@akhaliq Thanks a lot for sharing the updated version and for the effort you have put into the merge. I really appreciate the improvements so far. I have a few questions and suggestions regarding the current UI: 

1. The reference images/videos cannot be displayed properly inside the input box, and the input area is also too small for visual inputs.
2. For the understanding and editing tasks, the recommended case section currently only shows the prompt. This feels a bit insufficient, since it may not clearly demonstrate the expected visual input for these functions.
3. For the understanding task, the result is currently only displayed in a very small area in the upper-right corner. The output space is too limited to clearly show the understanding result. I would suggest centering this section and enlarging the output panel so the result can be viewed more clearly.
4. Could you move “Task FAMILY” out of the hidden model control panel and display it at the top center of the main interface instead? This would make all task types supported by Lance immediately visible to users, rather than requiring them to open the hidden controls. The overall layout could follow the current version for better visual consistency.
5. For visual generation, the generated result is currently displayed by filling the entire preview area. However, because the aspect ratio of the output may be different from the display area, the generated result may not be shown completely. For example, for a 1:1 generated image, if the preview area is wider than it is tall, part of the image may be cut off vertically. It would be great if this could be adjusted so that generated results can be fully displayed regardless of their aspect ratio.


Overall, I think these adjustments would make the interface clearer, more user-friendly, and more suitable for visual understanding, editing, and generation tasks.
I also want to sincerely say that the current improvements already make the interface look much more polished, elegant, and visually appealing than before. The overall UI feels cleaner and more professional, and it is clear that a lot of thoughtful work has gone into this merge. I really appreciate your excellent work and attention to detail here. Thanks again for spending so much time and effort improving the Space — the progress is genuinely impressive.

@ffy2000 thanks for the feedback, can you check now?

@akhaliq Thanks a lot for the super quick update! I really appreciate the effort you have put into improving the UI — the overall interface already looks much cleaner and more polished than before.

I checked the latest version, and there are still a few minor issues:

1. The reference images/videos or other visual inputs still cannot be displayed properly inside the input box. Also, the input area is still a bit too small for visual inputs, since images/videos usually need a larger preview space to be inspected clearly.

2. For the understanding and editing tasks, the recommended case section now includes visual inputs, which is great. However, the visual preview boxes are still quite small, making it difficult to clearly view the example inputs.

3. The overall display background seems to be black. I am wondering whether it might look more visually pleasant if we keep the original image background, or simply use a white/light background instead.

That said, I think the current online version is already good enough for a demo. The Gradio server indeed looks nice, but further polishing the interaction logic and visual details may take additional time and iterations. So please do not feel too much pressure to keep refining everything immediately.

Thanks again for your great work and for spending so much time on this. I really appreciate your patience and effort!