Positive users welcome Meta AI's interleaved images in text prompts as a long-awaited feature enabling new creativity, while negative users call examples unimpressive or cursed.
Based on 6 visible X reactions from 10 accounts; directional sample.
Ask a question below.
Published answers will appear here.
@giffmana I am wondering how to build an elegant text-image interleaved dataset. It seems that we don’t have such dataset in the real world? If the ability in the image is doing well, it’ll be an interesting thing to change the way people talk. Maybe IM apps will add this feature too lol
@giffmana Well idk where that leads to, but that's something new. I kinda waited this feature for 3 years from every model 😅
@giffmana This works very well with open-weight Gemma 4 models too. In that sense your screenshot is not so impressive
would love to see some @goodside creativity with this feature!
Something where I believe(?) MetaAI as a whole is currently leading, though I'm sure the others will catch up eventually:
@ezyang It would be misleading if the UI gave the user the impression that they are interleaved but the model just gets img1 img2 img3 prompt So for it to be useful this is not just a UI thing.
Positive users welcome Meta AI's interleaved images in text prompts as a long-awaited feature enabling new creativity, while negative users call examples unimpressive or cursed.
Based on 6 visible X reactions from 10 accounts; directional sample.
Ask a question below.
Published answers will appear here.
Meta AI's text input now supports direct images interleaved with text and I really like it! I always found it weird that other LLM UIs always had the pictures attached separately like attachments.