mov

Oh man, hiring is so broken, so basically:

"we want to hire someone to work in this super secret feature that uses lib X"

"cool! I'm the author of lib X"

"no, thanks"

That's a super interesting article and point of view. So, in your opinion, what's missing?

Thank you so much for the feedback! Already sent the message to our web dev team!

Far from vaporware, we're more like a steam machine full power mode ;-)

Hey hey! Thank you so much, we're really glad that you nailed the goal here: generative AI sometimes can go for a bit weird paths, but in our opinion it should first try to automate or augment what is pretty time consuming (and boring) for humans to do.

Please keep following us and drop us a line if you have any questions or suggestions!

That's true! I remember that!

Haha yeah, that's getting trendy. But to be honest, this style reminds me of the 90's like you said and it matches perfectly the street art and retro style you'll find in Brick Lane, where we have our London office: http://www.sundayupmarket.co.uk/

One of our co-founders and Artist Director is actually a street artist, so many inspiration from there :-)

Hello! So sorry to hear that our pricing model doesn't match your expectations. What would be an interesting pricing model in your opinion? Thanks for your feedback!

Hey! We can't share much what details of our ML pipeline for now but I can say that's a good guess :-) And you're right, we have bipeds, quadrupes, flying animals, vehicles, etc and we can actually handle any arbitrary model that belongs to one of those categories.

You can check our tool here at https://app.anything.world/animation-rigging

Thank you so much! We rig and animate for a general skeleton, but we have Unity and Unreal plugins that add controllers to our 3D models to make it easy to add the behaviour you want.

I'm from the AI team, so my Unreal knowledge has a limit :-D But please, join our Discord at https://discord.gg/anythingworld and I'll be more than happy to connect you with one of our Unreal Engineers to give you more details.

Hello! Thank you, glad you liked it! I'm from the AI Team but I already sent your question to our web dev team and they will answer to you pretty soon.

Hey hey! Thank you! Godot supports importing GLTF and other formats that we export, so you can basically use our tool (https://app.anything.world/animation-rigging), export a GLTF and import it in Godot.

Right now we only have plugins for Unity and Unreal, so you would have to create your own controllers in Godot. But having the rigging/animation already gives you a good start!

And please, share the details of your game project with us, we love to support indie game devs!

Do you mean https://app.anything.world/animation-rigging ? Please feel free to try it and let us know what you think of it!

Hello! Yes, when compared to Mixamo, our pipeline is not procedural nor static. We built an end-to-end ML pipeline that analyses the mesh/texture you upload and generate unique rigs and animations for your 3D model.

Different from Mixamo, we support all kinds of 3D object categories (animals, insects, vehicles, humanoids, etc) while Mixamo only provides a static gallery of already animated humanoids.

We just launched our autorigging tool today, please feel free to try it and provide your feedback: https://www.producthunt.com/posts/animate-anything

Thank you so much for the feedback! I already sent your comment to our web dev team!

Hello! One of the ML Engineers of Anything World here. I'm really glad that you appreciated the tool and I trully hope that it can be part of your stack and help you on your design.

We can't say all the details about what's under the hood, but you guessed it right. We indeed use ML models that we trained. It's part of a multi-step end-to-end ML pipeline that analyses not only the original mesh you upload, but also the texture and even semantics from the object it represents. Based on that we generate rigging and animations using also proprietary ML models.

Like you said, it's not perfect, but our ML models are learning every day and we just started! I really invite you to use the tool and see that it can be used as a starting point even by skilled artists to have an initial sketch and work from it.

Hope to have helped on your questions, and if you need more details, please feel free to reach out to us!

Hi, one of the ML Engs here. Yes you're right, we have lots of non-humanoids, which is something that Mixamo has been missing for a long time! But we also support humanoids. You can try yourself right now (we just launched our autorigging tool) and give us your feedback, we would love to hear it: https://app.anything.world/animation-rigging

People plugging it as output of Midjourney in 3, 2, 1...

SICP cover: https://upload.wikimedia.org/wikipedia/commons/9/9d/SICP_cov...

I would like to see more discussion around P2P solutions like https://dat-ecosystem.org/ and the fundamental principles around distributed networks/economies.

I really like the idea of being able to "git clone" your data/website into your own local browser and just make it's own webserver. For most of content we generate, that would probably be enough.

I don't see the point of trying to place everything in a blockchain. I understand the power (and need) to decentralization though.

And I'm not sure just one tech/coin/protocol/chain will be enough, but maybe if we pack:

- a P2P browser, with a wallet for authentication+payment (abstracting money away together with authentication is a near idea)

- a versioned content manager that make it easy to pull/push from/to peers

- ways to P2P nodes to sign their own content and give permission or not to peers to read/write, so we can trace ownership through version logs

- a way to run serverless functions to share computational power, but only for complex tasks like ML training, rendering, etc

- a way to automatically (and collectively) parse in-browser data to add annotations (bringing semantics to random data, or labels to ML datasets)

- automate renderers of such data to create (and host) 2D websites, data visualizations, or 3D content (you can call it your own miniverse)

Or what I just described is the Web we're already building, we just need more hands on deck to actually build it instead of wasting our time with bs.

Could you elaborate more?

Yes, throughput would be larger and we would loose the backprop path like you said, but it seems practical in some ways and actually guiding approaches like https://nv-tlabs.github.io/meta-sim/

Got it. So let's suppose we have an external renderer, then we could learn parameters to tweak rendered scene and then get rendered pixels, so we can calculate pixelwise errors from it and some target image we're trying to optimize for. In this way, do we still need differentiable renderer in your opinion?

Update:

It would require more training cycles and would not be as "atomic" as iterative tweaks but seems possible.

At same time, I wonder about making loss function talk with some external renderer would make it possible to mix both approaches.

So, in that case, we could learn without the need of a differentiable renderer neither graphics operators, right? Maybe the throughput of communicating with external renderer is too much when compared with iterative parameter tweaking happening inside of loss functions on CNN?

Very interesting report on your discoveries. Makes me remember of Genetic Programming. Was thinking about using the same principles to generate a more declarative Bayesian program (like a subset of SVG).

Do you have some source available around your experiments?

Back there I followed a bit the discussion and your alternative spec, which was really interesting. If I remember it well, that will take lots of work and you were the only one working to get it implemented on FF. Are there any plans to get back to that? Maybe as an independent API for JS sample processing by workers only, in parallel with WA? Congratulations on your past efforts and thanks in advance for your answers.

I do pretty much the same but using mosh instead of ssh. It's really stable even on low bandwidth of crappy mobile connection.

We also used sharp as part of a flow based pipeline for image analysis powered by NoFlo/Flowhub: https://github.com/noflo/noflo-sharp

We have been processing >3k images/day and >60k images/month.

An implementation of TodoMVC is really a good idea! There is already a (re)implementation of Jekyll using NoFlo that you might find interesting: http://bergie.iki.fi/blog/noflo-jekyll/

IRIS Explorer seems really interesting! I agree with you about that "loose of control feeling". I think it is a common effect that happens every time we go to a higher level or alternative paradigm. We have to create ways to easy go trough those levels and back again.

With Flowhub you can always go inside the 'black boxes' creating your own components as reported in the "Component Editor" section of http://bergie.iki.fi/blog/flowhub-beta. It is also possible to create and remix subgraphs that is another way to abstract a complex group of components.