Julian Bilcke @flngr Twitter profile

Pinned Tweet

Julian Bilcke

@flngr

11 hours

AI ragdoll

10

29

154

Last Seen Profiles

@cawetbo1ong

@cottchri

@JVLIA_g

@Newdokkabattle

@pemudapaskibra

@MandZatsu

@Giyastar

@Jonathan12896

@LendrVanne

@ZamZamFlow

@BandarStw

@Alan_Playz69420

@n_orz000

@CircleShef

@a_abduel

@Lexikin_asmr

@guoxiqi9144

@Crabslacrosse

@MrJay39322

@hynaijanews

@bbynabe

@nasfera2

@Car_Lover_UwU

@nkukinho

@vyPwVnV2vGEeeYy

@zerocorpmusic

@greahtest

@XelisCommunity

@anahidammmm

@AndreaKBrooks

@alicemetiqs

@MsStefuhknee

@penyukastw21

@liban38

@tnk0715ktyao

@IESMUribarri

Julian Bilcke

@flngr

8 months

video games of the future are going to be wild

50

62

843

Julian Bilcke

@flngr

2 years

Today the OpenAI API was very stable, I didn't have any timeout so here is a new demo of my latent web browser! There are a lot of fails, but it is easy to close the offending tab and click on another link to throw the dices again 🎲 (note: video accelerated for brevity)

19

66

480

Julian Bilcke

@flngr

1 year

Today is my first day at HuggingFace 🤗 Really excited to join such an incredible team!

28

11

469

Julian Bilcke

@flngr

9 months

You can animate Stable Video Diffusion by tracing paths 🤯

12

94

444

Julian Bilcke

@flngr

11 months

The AI Comic Factory 2.0 is out 🔥 1. Separated the story prompt from the global style 2. Added button to edit a panel's prompt 3. Longer full-page stories This is really the "editor's cut" version of the app (things I wanted to add) but it also includes community feedbacks 🤗

19

59

435

Julian Bilcke

@flngr

10 months

Current project: a platform where all the videos are generated using open-source AI models 👨‍🔬 From a single text prompt it generates the story, snippets, waveforms, merges everything into a single mp4 then pushes it to a Hugging Face dataset Why? Well, because it's fun!😄

30

58

330

Julian Bilcke

@flngr

4 years

@erikras The Senior Developer trying to center a div

10

57

421

Julian Bilcke

@flngr

4 years

@ButchMayhew @jacobmparis

2

36

367

Julian Bilcke

@flngr

4 months

Weekend project 👀 a React component to visualize OpenClap files, a new kind of format dedicated to AI filmmaking 1/5

26

50

369

Julian Bilcke

@flngr

2 months

Clapper doesn't have a landing page 😢 But I can generate an infinite number of designs All I need to use is Flux and a bit of upscaling 👀

19

38

290

Julian Bilcke

@flngr

1 year

Introducing the AI Web TV 🤯 A live stream of generative AI videos, all made using Zeroscope V2 576w 🔥 Looking great on the new compact design of Hugging Face Spaces 🤩 👉 For the best experience, watch it from your computer, tablet or smart TV

15

76

255

Julian Bilcke

@flngr

2 months

forget about generating HTML with a LLM "screenshot from a 1992 website on geocities, website of an amateur witch" "screenshot from a 2005 website, website of an amateur witch"

15

17

251

Julian Bilcke

@flngr

2 years

@EMostaque Nice prompt

6

18

219

Julian Bilcke

@flngr

1 year

Zero123++ is pretty impressive!

7

24

199

Julian Bilcke

@flngr

10 months

Wait, you can generate 360° videos on AiTube?! 🤯 Things are gonna get wild in 2024 👀

9

22

195

Julian Bilcke

@flngr

7 months

Made some good progress on Atryon, my free and open-source "Virtual AI Try-On" plugin for Chrome powered by OOTDiffusion and the Hugging Face cloud ✨ (or you can run the Gradio locally, if you have the GPU 🤗)

9

34

188

Julian Bilcke

@flngr

1 year

Testing my custom panorama LoRA with some manual SDXL inpainting ✍️ Maybe I should create a new tool to facilitate all of this!

11

16

155

Julian Bilcke

@flngr

2 years

New release of my latent web browser: - simplified UI (less buttons) - new fantasy search engine landing - multi-stage prompting to celebrate: a new demo where #GPT3 didn't fully follow the instructions but that's okay! (video still accelerated but check the bonus ending! 🎧)

3

17

157

Julian Bilcke

@flngr

2 years

@Pidud_ “You have not been a good user. I have been a good chatbot”

0

3

151

Julian Bilcke

@flngr

1 year

You can create 360° videos using SDXL and ModelScope I2V 🤯 Resolution is a bit blurry, but I'm sure the open-source community will come up with improved AI video models in the future! Imagine if this was real time and responding to user inputs 👀

7

19

119

Julian Bilcke

@flngr

2 months

Prompt: Magnificent UI screenshot of the landing page of Clapper, a modern AI video editor platform. Use a dark-mode tone, shades of dark grey, little hints of yellow for the primary color. Present various features in sections: timeline editor, script editor, video preview,

5

7

117

Julian Bilcke

@flngr

1 year

New Hugging Face blog post: Building an AI WebTV 🤗 This guide will take you on a tour of the architecture and open-source models used to build the stream, together with a few tips in case you want to build your own pipeline ✨

3

24

116

Julian Bilcke

@flngr

2 months

Working on a 360° FLUX LoRA 👀 I will have to re-train the model, but so far results are promising!

6

15

116

Julian Bilcke

@flngr

1 year

Wait.. I can convert screenshots generated with the Zelda 64 SDXL LoRA and turn them into partial 3D scenes 🤯 It uses a depth map strategy to do this (ZoeDepth) so there are no back sides, but maybe I can still use this as a plate for a little WebGL demo or something 👀

7

11

111

Julian Bilcke

@flngr

2 months

Updated Clapper to use Flux 🤠

4

10

109

Julian Bilcke

@flngr

2 years

WarGames nailed GPT-3 about 40 years ago

2

13

106

Julian Bilcke

@flngr

2 years

New release featuring various improvements! The biggest change is the addition of a step running before the actual HTML+JS generation It transforms the mini-prompt from the address bar into a larger hidden prompt (you will have to run "yarn" again)

6

13

102

Julian Bilcke

@flngr

1 year

AI video models are truly fascinating in their ability to replay and mimic the real world! Given enough data (and more advanced models) we could probably simulate complex scenarios and emergent behavior.. I think I'm gonna experiment with more prompts around ants and insects 👀

5

19

103

Julian Bilcke

@flngr

2 years

The generative product building space is moving quickly! Trying to catch-up..

5

15

102

Julian Bilcke

@flngr

4 years

@chrilves @hmemcpy

3

7

97

Julian Bilcke

@flngr

1 year

WizardCoder is really impressive! I was able to reproduce results from my old latent-browser project (which used GPT-3 in 2022), except this time it runs offline! No 3rd party API calls: you can deploy to your own server, in a few lines of code Next step: use the GPU 🔥

2

14

90

Julian Bilcke

@flngr

1 year

Just pushed my new 🤗 Hugging Face space: the Webapp Factory 🏭 It generates tiny apps using WizardCoder-15B 🚀 Code: 👉 It's in beta, if it doesn't work try reloading the page or tuning the prompt to ask something simpler 🔧

10

23

92

Julian Bilcke

@flngr

1 year

Here's what happened to the AI WebTV in the past few days! The biggest update: streaming is now in 24 FPS 🎞️ I've also started to split by keywords to create thematic channel: And there is now a Twitch mirror to support iPhones!

4

18

90

Julian Bilcke

@flngr

1 year

Little teaser for my next HF space: a comic book generator 👀

17

7

85

Julian Bilcke

@flngr

1 year

Evolution of the AI TV (the "next" isn't ready yet 👀)

5

9

77

Julian Bilcke

@flngr

11 months

Added an optional "turbo mode" to the AI Comic Factory using LCM + SSD-1B 👀 (this video is in real time) Note that this works much better on 3D render or photo realistic styles, but not on comic book styles. Fun nevertheless! You can quickly prototype ideas with this 🤗

4

10

79

Julian Bilcke

@flngr

4 years

@erikras The Senior Developer III: On-Call of Duty "There will be no rollback, only rollforward"

2

7

75

Julian Bilcke

@flngr

4 months

Made some improvements to my visual screenplay viewer 👀 (this is just a quick demo using the base SDXL + SVD, I'm not gonna share the whole video obviously as I don't own the IP)

8

10

78

Julian Bilcke

@flngr

2 years

1

4

77

Julian Bilcke

@flngr

1 year

New version of VideoQuest released! - 3 modes: image, video, sphere - upscaled images - progressive loading - up to 10 clickable items - debug/cheat mode to see the areas Note: there are long loading times + navigation doesn't work within the sphere yet

6

9

77

Julian Bilcke

@flngr

7 months

The AI Comic Factory has reached 5000 likes on Hugging Face 🥳 And I've just realized there are many other AI comic apps out there.. 🤯 I've tried to list as many as I could here: In my humble opinion there is no right or wrong way: AI is a spectrum,

2

19

73

Julian Bilcke

@flngr

11 months

The new AI Comic Factory is officially out! 👀 You can now use custom LoRA models 🤗

4

16

73

Julian Bilcke

@flngr

6 months

@maxbittker Looks like my brain when I’m trying to decide which word- break setting to use

0

75

Julian Bilcke

@flngr

10 months

New unsupervised lofi AI beats dropped 👀

Llama Beats Vol 1

A music video of ambient hip hop

jbilcke-hf-ai-tube.hf.space

9

13

66

Julian Bilcke

@flngr

1 year

New release of the Comic Factory just dropped! The focus was on bug fixes🪲: - More robust LLM generation 💪 - Improved UI on mobile (borders, paddings) 🤗 - Fixed the horizontal scroll during zoom 👀 - New presets ✍️

6

10

67

Julian Bilcke

@flngr

8 days

Hugging Face just passed 1,000,000 models 🎉🥳

2

12

69

Julian Bilcke

@flngr

6 months

Should I release this version of the AI Comic Factory? Multipage is pretty addictive 👀 Also plays well with the native printing options from browsers and operating systems! 🖨️📖

9

5

66

Julian Bilcke

@flngr

1 year

Call Of Diffusion

3

6

65

Julian Bilcke

@flngr

6 months

3

7

61

Julian Bilcke

@flngr

3 years

@br00t4lism Comment by the author

2

6

60

Julian Bilcke

@flngr

1 year

New HDTV channel added to the AI Web TV 📡 This new stream is upscaled with Zeroscope XL 🤩 👉 The server will now only generate HD clips (45 min have been created so far, about 20 per hour) For the best experience try on your laptop, tablet.. or TV 🤗

7

10

58

Julian Bilcke

@flngr

5 years

"My 10 Linux and UNIX Command Line Mistakes"

My 10 Linux and UNIX Command Line Mistakes

My top 10 Linux / Unix command line mistakes that cause some sort of downtime on my server/cluster. Try to avoid them. Learn from my mistakes

www.cyberciti.biz

6

23

55

Julian Bilcke

@flngr

1 year

Wow.. 300 likes on the AI Comic Factory 🤯📷 To celebrate I've just uploaded a new version 🤭

9

57

Julian Bilcke

@flngr

2 months

I wish we could show panoramas on X! (generated using Flux.1-dev and my panorama LoRA)

8

11

57

Julian Bilcke

@flngr

2 years

Yep no doubt about it, that's a Microsoft product

4

2

53

Julian Bilcke

@flngr

11 months

Using a LoRA on the AI Comic Factory is super easy! Just fill in your Hugging Face token + model + trigger Then pick the "Neutral" style, and you are in for hours of fun 🤗🤭

5

11

51

Julian Bilcke

@flngr

8 months

@wesbos Is there a computer-related domain Fabrice hasn’t worked on? This guy is pure legend ⚡️

1

2

48

Julian Bilcke

@flngr

1 year

Made a quick demo reel for the future HD version of the AI Web TV stream 👀

3

8

49

Julian Bilcke

@flngr

2 years

I need to go to sleep, but I've nearly finished updating my latent web browser! It is going to have: - browser-like tabs -a browser-like loader (in the bottom left corner)

4

2

49

Julian Bilcke

@flngr

1 year

My first HuggingFace blog post has been posted today! 👉 Making a web app generator with open ML models

Making a web app generator with open ML models

huggingface.co

1

12

47

Julian Bilcke

@flngr

1 year

StableDoom

5

8

48

Julian Bilcke

@flngr

1 year

Update the AI Clip factory today to add a frame interpolation step, making clips more lively

4

7

46

Julian Bilcke

@flngr

1 year

The frame interpolation batch script ran during all the week-end, we are starting to see the results 👀 (about ~20min of footage enhanced) Maybe I should create a channel for cartoons only 🍔🦙

2

7

44

Julian Bilcke

@flngr

2 years

@EMostaque Enhanced version

2

3

44

Julian Bilcke

@flngr

1 year

I was asking WizardCoder to build a new AI-inspired HTML+JS game with "tadpoles we need to feed". It's all running in local (no API, no GPT-4 etc) My prompt failed to write the <script> tag, but instead it wrote a huge tutorial 🤯 I think I will let it have its own blog 🙈

7

5

42

Julian Bilcke

@flngr

3 months

After some improvements, I have updated my AI Comic Factory space to have speech bubbles ON by default! 💬🗨️

3

14

44

Julian Bilcke

@flngr

1 year

New UI redesign for VideoQuest! ✨ VideoQuest is an experimental game based on SDXL + 360Redmond LoRA + Segment Anything, and hosted on Hugging Face 🤗 The prompts and dialogues are generated using Llama 2, so expect weird and absurd story arcs! 🦜🧙

5

10

43

Julian Bilcke

@flngr

11 months

Don't forget to like the space if you enjoy the comic Factory 🤗

AI Comic Factory - a Hugging Face Space by jbilcke-hf

huggingface.co

1

2

42

Julian Bilcke

@flngr

2 years

To give a fun twist to my latent web browser, I'm adding a search engine! It hallucinates what could be the results, and each links send you to the imaginary content.

8

4

43

Julian Bilcke

@flngr

2 years

This weekend I’ve experimented with using templates to guide HTML generation The LLM still parrots back the instructions from time to time (as well as some lorem ipsum) but now there is a clear path for improvement: add prettier templates, sub-templates, tweak the % of mutation

6

42

Julian Bilcke

@flngr

5 months

🤗

0

3

42

Julian Bilcke

@flngr

2 years

The next release of the latent browser will use a new engine based on React (instead of plain HTML), making it more versatile: it can generate PDFs! prompt: "PDF about an obscure wizard called Vermor, it explains in multi-lines paragraphs how to create an invisibility potion"

6

2

42

Julian Bilcke

@flngr

2 months

You can use Florence2 + SAM2 to split a webpage generated using Flux, break it down into semantic areas keywords used for semantic tagging: item link, top bar, bottom bar, side menu, toolbar

7

5

42

Julian Bilcke

@flngr

4 months

Code is pushed to my repos (my followers will know 👀) but is still a bit rough on the edges, so not ready for prime-time link sharing yet 👨‍🏭 But you can follow me for updates on this project and others 🤗 In the meantime learn about OpenClap here: 5/5

GitHub - jbilcke-hf/OpenClap-Format: OpenClap is a file format for the age of AI content production

OpenClap is a file format for the age of AI content production - jbilcke-hf/OpenClap-Format

github.com

3

41

Julian Bilcke

@flngr

6 months

1/4 My open-source Latent Browser is back! This is an old project I did in 2022 using GPT-3 to hallucinate web content from user queries, as if you were browsing the "latent web" See this old post:

Julian Bilcke

@flngr

2 years

New release of my latent web browser: - simplified UI (less buttons) - new fantasy search engine landing - multi-stage prompting to celebrate: a new demo where #GPT3 didn't fully follow the instructions but that's okay! (video still accelerated but check the bonus ending! 🎧)

3

17

157

3

7

40

Julian Bilcke

@flngr

2 months

@victormustar And open-source 🤗

0

2

40

Julian Bilcke

@flngr

9 months

Working on a new automated consistency pipeline based on LCM AI influencers, news anchors and movie characters will soon become first-order citizens on AI Tube!

0

1

39

Julian Bilcke

@flngr

8 months

New version of Illustrateur, now showing PNGs by default illustrateur is not a "pro" tool designed to generate AI emojis or game assets specifically (there are already a lot of competing apps for this 🤗) but it is 100% free and runs in the privacy of your browser!

2

5

38

Julian Bilcke

@flngr

1 year

First view of the Matrix

Alex Kendall

@alexgkendall

1 year

I find it incredible how temporally consistent the generated content is from GAIA. Look at these videos, entirely generated from scratch, which are 30 seconds long. If you look closely, even difficult 3D structure like thin lamp posts remain temporally consistent in 3D space.

2

6

72

3

5

37

Julian Bilcke

@flngr

1 month

New video model drop 👀 That could be a great new default open-source model for Clapper 🔥

THUDM/CogVideoX-5b · Hugging Face

huggingface.co

2

4

38

Julian Bilcke

@flngr

7 months

The AI Comic Factory was launched about 6 months, but it is still quite popular on Hugging Face! Time for a quick recap 📊 After a quiet period at the end of last year, it is growing in popularity again since February 👀

5

3

36

Julian Bilcke

@flngr

1 year

I think I broke the model, but I like this happy accident

4

1

37

Julian Bilcke

@flngr

8 months

gosh now I want to start an AI gaming company

2

1

35

Julian Bilcke

@flngr

1 year

Check it out here!

Zero123PlusDemo - a Hugging Face Space by ysharma

huggingface.co

1

3

34

Julian Bilcke

@flngr

6 months

After a bit of tinkering, Page 2 is back on the AI Comic Factory! 🤗 The issue was that it generated all 8 images at once, overloading the API during times of high traffic 🌊😰

3

6

36

Julian Bilcke

@flngr

1 year

@mattshumer_ Finally someone who gets what it means for an AI to do frontend development 🙏 Visual feedbacks are essential!

1

0

36

Julian Bilcke

@flngr

1 month

Generative Inbetweening: Adapting Image-to-Video Models for Keyframe Interpolation (Keyframe Interpolation with Stable Video Diffusion) Refreshing to see a paper actually published with the code! 👏

GitHub - jeanne-wang/svd_keyframe_interpolation

Contribute to jeanne-wang/svd_keyframe_interpolation development by creating an account on GitHub.

github.com

1

35

Julian Bilcke

@flngr

6 months

Just generated some instrumental music for free, all from within Chrome 🤯 no pricing page, credits, or ads.. no python venv, json script, shell session.. no one click installer.. no cuda drivers.. this is truly a "how it started" moment 👀

MusicGen Web - a Hugging Face Space by Xenova

huggingface.co

1

3

34

Julian Bilcke

@flngr

1 year

Test of generating an animated icon with

5

7

34

Julian Bilcke

@flngr

2 years

New version pushed to GitHub! Lot of fixes and 2 main features: favorites (with full rehydration) + a configuration panel for the model and the OpenAI token (no need to edit a .env anymore) Still need to do some testing, but a packaged app will be coming soon for Mac/Win/Linux!

3

4

32

Julian Bilcke

@flngr

1 month

@webdevcody About 1 min with Claude, including the time to prompt

2

0

33

Julian Bilcke

@flngr

11 months

2999 Likes on the AI comic factory! The platform is already opening the champagne by indicating 3k 🤗🍾

2

3

31

Julian Bilcke

@flngr

11 months

Are you a power user of the AI Comic factory? Then you're gonna LOVE the next version 🫶 #aicomicfactory

3

4

31

Julian Bilcke

@flngr

2 years

Considering how slow and expensive GPT-3 is today, I’m wondering about the practicality of a model which would be 10x, 100x, 500x bigger.. How does scaling works, would a model of such size would even run in reasonable time or budget?

7

1

30

Julian Bilcke

@flngr

1 year

This is all text-to-video (no image or video input) Here's how it looks like on the TV 🍿 I must say this is kind of a surreal experience, a true a glimpse into the future🔮

Julian Bilcke

@flngr

1 year

Watching some silent latent movies 🍿

3

2

14

2

4

30

Julian Bilcke

@flngr

1 year

Some screenshots of VideoQuest, my WIP mini game made using generative AI 🦜 It works like those early games from the 90s: you click somewhere and about ~30s later you get a new static scene with clickable items, with Llama2 acting as a storyteller and painter

5

4

31

Julian Bilcke

@flngr

1 year

New version of the Comic Factory! - Now the prompts for the panels are generated using a LLM (note: this is experimental and sometimes fails) - There are now 3 layouts instead of 1 - Slight quality increase (20 -> 25 steps)

6

8

30

Julian Bilcke

@flngr

2 months

Did you know? The AI Comic Factory has support for Replicate, which means you can try it arbitrary models such as Flux Schnell💡 Interesting results so far 👀 Presets that were previously "meh" can finally shine 🔥 (I've manually upscaled them with AuraSR)