Flux 3

(bfl.ai)

142 points | by ThouYS 1 hour ago

14 comments

  • user43928 56 minutes ago
    I hope the open-weight versions will be SOTA.

    > Over the next few weeks and months, we will make the following capabilities available

    > Open-weight access to a multimodal backbone, for content creation (video, audio and image) and action prediction. (“FLUX 3 Dev”)

    > We will also release more technical details on the underlying approach.

    • 3form 13 minutes ago
      Open-weight promise seems nice, but 1) I've seen people commenting that these hopes amounted to nothing for some previous releases (no idea which promises were made though); and 2) if the backbone is released as Dev, what will be missing? I can't easily tell from the post.
      • musebox35 3 minutes ago
        dev variants are usually cfg distilled which means that directly finetuning isn’t as effective. In the past, for the flux2 klein models,they released base versions that are not distilled. So it will probably be a while before you can fully take advantage of the open weights.
      • iLoveOncall 3 minutes ago
        I run some of the 2.3 models locally so I'm not sure where you saw that they didn't follow their words?
  • thisisauserid 1 hour ago
    - Showed close to zero examples of people.

    - Frivolous use of the term World Model.

    - Claims 20 seconds of video, shows only jumpcuts.

    Coming soon!

  • saejox 2 minutes ago
    i don't get why they are investing money on image/video gen. All generations i have see looked blurry, lacking in fine details and missing the artistic touch (lacks meaning? lifeless?)
  • jdthedisciple 26 minutes ago
    It's incredible how negative and dismissive the comments in here are while here I am thinking the model actually looks impressively capable.

    But then again I heard the downers have always been the first to leave their dung comments here so let's see...

    • muppetman 22 minutes ago
      Advertisers love these sorts of reactions.

      I see nothing here but them TELLING us how great it is. Not showing us.

      • jdthedisciple 15 minutes ago
        > Not showing us

        Have you watched the 46s video full screen on a monitor and not marvelled at the incredible 4K detail of the FPV motorcycle racing clip?

    • ra 23 minutes ago
      It would be interesting to see some time-series sentiment analysis of HN. Subjectively it feels like there's much negativity on HN these days, I hope I'm wrong.
    • drsalt 16 minutes ago
      none of this is exciting compared to holodeck in star trek
  • Gecko4072 19 minutes ago
    I thought the clips were real footage until they were dancing in a flooded room.
  • zmmmmm 57 minutes ago
    > It jointly learns from images, videos, and audio within a unified architecture, because what it needs to learn is not any one of these elements in isolation.

    I'm confused, videos contain images and audio ...?

    • ibotty 43 minutes ago
      That's most likely a disagreement on terms. In the media world, video is only the moving images, not audio. This is separate from images, that are meant to be still images.
    • PxldLtd 40 minutes ago
      It's more a comment about the feature detection I think; all image, video and audio input contribute to the same weights/activations that can produce image, video and audio output.
  • SubiculumCode 52 minutes ago
    Well, unified multimodal intelligence is the only way we will get to The Terminator, which seems to be the goal now of Silicon Valley and every Nation State with a military budget, so have at.
    • UberFly 49 minutes ago
      I wish you were being hyperbolic but I know better.
      • SubiculumCode 30 minutes ago
        I don't even know anymore, to be honest. I talk to AIs more than I do with humans, these days...So who knows
  • teiferer 1 hour ago
    Lots of words about multi-modal but then this:

    > our mission to develop real-world visual intelligence

    Visual is mono-modal, isn't it?

    • nerdsniper 1 hour ago
      Is this really the value-add comment you’re going with?
      • camillomiller 58 minutes ago
        Says the one who posts this comment?
  • luciana1u 35 minutes ago
    imagine spending nine figures training a model to learn that the sound has to match the impact. my 8-month-old figured that out by dropping a spoon on the floor twice.
  • frotaur 59 minutes ago
    Sorry because pointing this is a bit tired by now, but reading already the first two paragraph thete is this unmistakable stench of LLM slop writing. Immediately disengaged.
  • mattmanser 57 minutes ago
    Open-weight plans are near the bottom (Launch section):

        - Video and audio generation and editing through APIs and private weight access. (“FLUX 3 Video”)
        - Action prediction through selected research and commercial partners, beginning with mimic robotics (“FLUX-mimic and FLUX 3 Action”)
        - Image synthesis and editing through APIs and private weight access. (“FLUX 3 Image”)
        - Open-weight access to a multimodal backbone, for content creation (video, audio and image) and action prediction. (“FLUX 3 Dev”)
  • vouaobrasil 50 minutes ago
    The fact that people keep developing this technology shows that the true problem is not that machines are likely to become intelligent, but that people have already become machines - unthinking and without any care to the future whatsoever.
  • doitright99 25 minutes ago
    AI slop trained on copyrighted content.
  • vladsiu 23 minutes ago
    [dead]