Creator of the Lil Yachty viral walk out trend

Canada 🇨🇦
Viggle AI Experiment 💙 This is an incredible tool for style transfer. Details and link to Discord below 👇 As always, if you enjoy this content please kindly like and share!
198
764
4,247
683,321
Here's probably one of the hardest shots I've ever attempted 1) Two characters 2) Two different race of humans (prompt bleeding) 3) Very fast paced with a lot of motion Was very fun and I learned a lot. Also develop a bunch of new workflow #StableDiffusion #Matrix #Pixar
892
485
4,409
8,641,248
ChatGPT prompted 74 times “Create the exact replica of this image, do not change a thing” This is why I say you need to start a new chat after each edit 🤣
208
213
2,668
574,455
💙Kendomland - land of the free and the men 🩷 This one was fun and feels unique I left the expressions and mouth movements more muted intentionally as it depicts a "doll" more clearly... I think? As always - please like and share if you enjoy this content! Love you all
92
396
2,554
262,358
Hunyuan vid2vid severance to John Wick Made by allhailthealgo Brilliant. One of the best I’ve seen
88
261
2,305
487,524
"we've got Harry Potter at home"
4
3
2,285
80,114
🚨I made a ComfyUI Wrapper for NormalCrafter (Video normals) It's just the one node, extremely minimal reqs.txt to install and it auto downloads all the models for you on the first run. Please give it a try and let me know if it works for you LINK IN FIRST COMMENT BELOW 👇
50
200
1,966
111,021
This is so good 🤘🏻 MJ Video + Banger track + creative idea
AoKi ¹⁰⁴
32
195
1,943
87,424
Another example here of SAM V2 with literally 1 mouse on the first frame of the video and it tracked Simone even through that super fast motion part at the end If I were a VFX shop I'd be implementing this asap. It's better than any traditional ROTO I've ever used
44
200
1,627
156,985
TENCENT casually dropping a 13bn parameter open sourced video model WITH weights 🩵 Looks unreal and can't wait to test it
35
190
1,542
156,365
Image to depth using Marigold and ConfyUI Then depth to 3D geometry in Blender Super interesting stuff 🤯
30
205
1,471
202,176
New SAM2 is an absolute game changer.... Did this track in 1 mouse click and about 4 secs. The fact that it even tracked across the scene change is mind blowing to me Will post some more tests below
26
153
1,489
171,942
How I made this 🧵 As discussed - if a post gets decent reception I’ll start doing a high level brain dump about the process. This post had above average feedback so I’ll talk through it below As always, please like and share if this is useful - it helps me. 👇🏻
39
117
1,377
344,984
👴For those asking - yes you can also go in the other direction (Old -> Young) 🧒 This was 364 frames @ 1920 x 1072 resolution. Note: As we know what the younger version of the actor should look like this task becomes more difficult to 100% nail the likeness with zero shot
64
120
1,341
705,425
Hard to find motivation to build anything right now when Nano Banana will just obliterate you in a week or two
60
71
1,310
108,698
🚨Release - WarpTuber (A real-time VTubing Tool) (link to download in first comment) 👇 I've included a premade VENV. In a perfect world you just unzip this and it works out of the box (tested on Windows 11) My hopes are that people smarter than me pick it up from here and run with it! More comments below 👇 $WARPER
30
146
1,138
96,422
Fun workflow I was playing with last night 1) Kontext to remove Thor from the shot 2) Photopea to place Shrek 3) Kontext + Relight lora to blend him into the shot 4) Wan2.2 i2V to animate Very addicting... 😍 Prompts are written up in the corner. Wan 2.2 prompt below 👇
38
89
1,138
95,887
Diffusion even set the mood to be more "jokeresque" !
27
160
1,060
126,324
Hedra's new "Character 3" audio 2 video system is utterly mind blowing to me. Video below was generated with a single image + an audio file (no prompt - I left it blank). If you can't see the potential of this - no hope for you.
50
113
1,092
113,150
FPS in top left corner 40fps realtime with video & 4.6gb VRAM ✅ 50fps realtime with image & 1.7gb VRAM ✅ Running locally on a 4090 o3mini w/ Cursor go brrrrrrrrrrrrrrrrrrrrrrrrrrrrrr
53
110
1,062
110,967
"Bro, it's easy just install ComfyUI"
63
44
1,068
75,244
The matrix in the style of Netflix Castlevania Proof of concept with Wan 2.1 and stylized first frames with RF_Edit Added some bloom in davinci resolve and upscaled with topaz but outside of that it's raw output.
50
109
992
80,196
This is so insanely good and would take so much effort to make without AI Anyone saying AI has no use case is truly a complete and utter imbecile
ComfyUI
561
138
1,701
127,216
Sora 2 “physics” 🫠
68
43
998
170,909
Bruh...... this is local on fully opensource tools! What a time to be alive..
28
92
977
85,002
Wan2.1 Vace Source left, output middle, ref frame right Most certainly a new frontier with these local video models... look at the light reflections on the jewels in the crown 👀 81 frames (every 2nd source frame) then interpolated 2x
31
91
965
69,368
2yrs of progress I should dig up some of my old renders and redo them too just to see how far we’ve come
Justine Moore
32
97
972
58,853
Experimental workflow test 1) Rough Blender Animation Blockout 2) #RunwayML Gen 1 Text to Prompt for Base Style 3) #AnimateDiff for cleanup Rendered at 8 fps for time saving (just a test at this point) Interesting results #AIVideo #AIAnimation #StableDiffusion
25
109
903
121,162
Somehow #adobe hasn't included gen fill into their After Effects product yet for videos... which BAFFLES me So I did it myself. Fully open source✅ All free tools✅ GG✅ Your move Adobe
47
76
887
96,850
You’re seeing an 11-12 figure business sector emerge right before your eyes I don’t see any scenario where this isn’t baked into every retail clothing vendor that sells online clothing This will be able to run on a TI 83 in a year 😂
Hirokazu Yokohara
40
76
876
174,085
If you thought the tok was brain rot…. Wait until you get access to this thing…
30
68
863
78,993
Luma AI released their "Reframe" feature today This can be thought of as outpainting in the traditional sense This was a promptless first attempt... VERY promising! The black box in the middle is the source video... everything outside of that is inferred by AI 🤯
33
92
808
79,708
My ADHD trying to keep up with all these tools be like
18
59
803
33,396
Immersion ✅ Clever idea to be honest.
RŌHKI
24
98
746
59,728
Object removal Wonder Dynamics vs my Comfy workflow Thoughts?
33
78
730
33,932
Wan 2.1 is the most fun I've had in a while. Style transfer feels soooooooooo close to being there.
17
65
734
36,214
👨‍🍳The longest continuous shot I have ever done🐻‍❄️ 2 clips: First: 909 frames in 12 context windows. Prompt executed in 00:39:25 Second: 2337 frames in 31 context windows. Prompt executed in 01:32:02 Max allocated VRAM=27.458 GB 720p resolution Need 260gb of System RAM though
25
72
734
56,917
I've been playing with some aging workflows to test the viability of these tools for more traditional VFX tasks. Pretty decent results and I suspect someone with an actual Nuke / Davinci background could really make it work. Thoughts?
74
58
724
79,816
Wan 2.1 (1.3bn) text-2-video Flat Color Lora Control Lora Depth 🥵🥵🥵🥵
21
86
707
31,601
Imagen v3 vs Imagen v4 thead: Prompts in alt. Left will be Imagen v3, right will be Imagen v4
38
33
638
123,202
Just came across this on Reddit by user Protector111 (link in first comment) This entire video is made with Hunyuan + custom Loras Seriously one of the best AI anime outputs I’ve ever seen 🥵
22
65
621
55,943
Wan 2.1 Westworld Lora Upscaled with Topaz Starlight The upscale worked quite well on this one.... I think?
24
58
617
55,047
I was able to achieve a decent full face DeepFake using only opensource tools that didn't require a DeepFace labs training set (No model training at all actually): ComfyUI EBSynth 100,000 views and I will share the .json and explain the process (it costs you nothing to share)💙
35
108
621
167,657
NGL…. New Rodin 1.5 looks unreal 2025 is going to be the year where image to mesh significantly improves. This is already pretty wild Will test and post some examples… let me know if you have an image you want tested
モジョン
19
62
610
47,430
Created with the free object removal workflow I shared 🛹
Replying to @AIWarper
Thanks, pretty impressive and pretty fast.
13
53
621
50,154
Update on my experiments with a consistent full style transfer - the fast movements with blur are killing me. "toony / Pixaresque" look to Keanu. Things of note: 1) Belt changes color (I'd fix that in AE in post) 2) Some minor eye jitter at the end (fix in post) Fun stuff!
33
62
584
76,885
Just wow..... absolutely amazing Balloon scene from up. Video desaturated so you can see the balloon I am tracking with SAM v2
20
55
577
29,742
What if you could use AI as a Render Engine? 🤯 Check out this video where I replace Captain America with someone you may know 😉 (🔊 on) If you enjoy this content please consider liking and sharing - it really helps me grow! As always, details below 👇
15
59
538
89,912
Some early testing of this opensourced "Act-1" There are so many knobs to turn and test but it's incredibly promising I can get it better still this is just a couple hours of tickering - I want 1:1 expression transfer. Stay tuned!
27
64
556
57,820
Full video with sound! This is by far the best transfer of this particular video i've been able to acheive. It took: 4 different ComfyUI workflows A stack of nodes in Davinci to clean it A topaz Upscale pass Conclusion - gated to the commoner lol
24
42
516
88,308
Local open sourced tooling is just so much more fun than closed source. No safety guard rails No censored prompts No canceled generations due to copyright infringement You can just blast
Transfer character with VACE for Wan 14b. Generation by @ingi_erlingsson
12
57
543
24,043
"No more forks?" Some thoughts in the comments 👇 Please kindly like an reshare if you enjoy this content it really helps me in the long run🩵
35
62
518
47,839
Slopfactory open for business 🧌. Created using WAN 2.1 + Flux Kontext + Flux Fill + ComfyUI Workflow and breakdown to follow. Known issues: Lipsync could be better I could have composited the original woman back Start to finish ~1.5hrs of work
7
63
530
49,881
Wan 2.1 + VACE vs ACT two I admittedly did not spend much time on this could very likely get it better
18
51
522
37,011
For anyone curious what my last tweet was about - I made this almost exactly a year ago today using only: ComfyUI DaVinci Resolve Curious to see how much better we could make it now a year later
A.I.Warper
27
43
513
41,376
Kijai converted the CausVideo into a lora. The TL:DR here is this is just 4 steps, 81 frames, 576 x 1024 Prompt executed in 198.92 seconds To put that into perspective, my previous example of this (using 1.3bn VACE) took 10-15mins to do 81 frames. MORE DETAILS BELOW 👇
19
56
513
40,871
Image 1: Generated with my Influencer Lora Image 2: Downloaded from ALO website Image 3: Virtual try on + upscale Conclusion? I can very likely make any ad for any clothing company in less than 5 minutes
23
42
498
84,821
I'm sure X will compress the heck out of this but..... Incredible detail coming out of this FLUX model. Truly incredible with only a second ksampelr pass for upscale and nothing else
21
54
496
39,246
Can’t stop won’t stop 🥵🥵🥵🥵🥵🥵🥵🥵🥵🥵
33
39
492
52,491
🚨My base WAN 2.1 14bn vid2vid workflow is now live, for free, on my Patreon. LINK IN FIRST COMMENT BELOW 👇 81 frames at 576x1024 with 4 steps Max allocated memory: max_memory=10.090 GB 14bn model handles motion much better than 1.3bn
26
65
500
43,395
Generated locally on a 4090 with Mochi + MochiEdit custom nodes New video model arc is beginning and I’m here for it
24
54
491
42,335
If you were waiting for MatAnyone to be wrapped into Comfy the wait is over! This tool is incredible and fast. I was getting my outputs within 30s on 5s clips. Link to report in first comment below 👇
Purz.ai
14
66
489
46,416
I've gotten DMs asking how these viral "behind the scenes" of Kpop Demon Hunters Live Action AI videos were made. It's quite simple actually - here's how 👇 If this helps you please kindly like and share 💙
29
49
490
69,598
How to do this: 1) generate a still image of the girl 2) film yourself pretending to be real time with tame movements 3) Pipe through live portrait 4) upscale output with topaz 5) composite your source video from step 2 down on the corner Profit
ユニコ🦄 ClaudeCodeセミナー11万人の人🔥
17
62
480
49,739
Fixed the proportions on my Blender rig to not look like a skinny alien👽 Much better results! Only thing left to try would be to fine tune the animation to look more 2D hand drawn (animated on 2s) 🩵If you like these updates please like and share! It helps me grow 🩵
17
48
465
76,952
Tencent must have gotten into my god awful stash of test videos I use This looks sooooooooo promising.. very excited
24
36
473
85,897
Fully Automated object removal with SAM V2 Still at the mercy of the existing object removal models but when someone trains a good version of that it's lights out Mask didn't include smoke / car shadow so obviously it remains Not bad for a few mouse clicks
14
50
476
24,679
LUMA AI have created a state of the art tool with one of the lowest barrier to entries in the market. Paired with Flux Kontext, it's never been easier do film yourself and do a complete style shift to something else. It's also addicting af to just mess around with
19
55
455
21,946
I mean come on......... Minus the first couple frames ghosting the quality of this is actually insane
37
28
457
30,519
And y’all thought we’d have to pay Runway for stable vid 2 vid This is fully local Fully open source Fully FREE TO USE (Video created by 👑 Kijai)
33
47
445
34,827
Kijai the🐐 threw together a ComfyUI implementation 😊 github.com/kijai/ComfyUI-cha…
9
81
442
31,991
HelloMeme released an update to their Comfy code (v2) Pretty good imo... arguably best in class for this use case. Some minor wobble in the background but overall pretty good... inference took about 10 mins for 250 frames on a 4090 Repo below 👇
12
63
438
34,776
WanAnimate tests in ComfyUI 121 frames at 720p.. took ~60gb of VRAM to run to avoid context windows🤣 Minus the eye direction... quality is crazy good
31
33
444
53,501
How on earth is this entirely free to run locally on my computer. None of these closed sourced tools can get that laughter correct This is non cherry picked too I could have fished for a better variant Thoughts?
26
40
429
40,839
Replying to @djcows
Must be @elonmusk ‘s California king in the corner?
7
3
414
76,454
lil taste of what I've been working on Massive improvements to my style transfer with expression and lip sync carry over Please be sure to like and share as it helps me grow! #Animatediff #animation3d #AI #Stablediffusion
28
40
418
27,337
FLUX Redux with my RUBBERHOSE Lora Works pretty well! (link in first reply)
18
29
426
49,728
Remove the soccer ball and you are left with an oddly funny video 😂 Fully automated object removal with SAM V2 + ProPainter
6
44
421
57,891
It's never been easier to make an AI powered influencer now with FLUX 🧵 FAL just introduced an insane LORA training tool that is actually insane. Meet Mei Trix (see what I did there?) READ BELOW ON HOW I MADE HER WITH ONLY 4 IMAGES 👇
16
40
395
65,735
Elon dropping major alpha on the latest Rogan podcast It's so insanely simple to make anyone say anything you want these days... these Elon clips especially feels super good. Entire thing took 20 mins....care to know how?
41
20
418
177,036
🚨MODELS RELEASE🚨 🔹WAN2.1 VACE-1.3BN (was previously preview only) 🔸WAN2.1 VACE-14BN Interesting that they're now just modules that bolt on to the exist models (any of them) LINKS BELOW 👇
10
61
409
22,234
Far from perfect but done completely with local open source tooling on a 4090... for "free" 1) Make a ref image using FLUX 2) Use LTX video 2 video with the nodes below This video model is smaller than SDXL.... wait until we get 5b-10b variant and it's GG
32
53
396
39,451
"You'll have to talk louder, son!" Experiments with lipsync + expression transfer to a different medium. Some thoughts (Continued below as well) 1) I'm at the mercy of the available checkpoints. The "Disney / Pixar" models for 1.5 are very meh
23
43
385
42,177
You know what time it is..................🤫🤫🤫 324 frames, 720 x 720 Prompt executed in 00:19:15 (large portion of this was 2x interpolation with RIFE) Very happy with how this turned out... great new tool in kit 🩵
13
27
398
49,194
Eye gaze fully fixed in Comfy now for Wan Animate. All you need to do is use the pre processing nodes I linked in the first comment below for your pose / face crop. Tested this against Wan's Native source code and it is indeed 1:1 functioning now this was 300+ frames
22
30
394
36,528
LTX vid2vid not to be slept on either Crazy thing here is this is 149 frames and the entire render took 20 seconds locally on a 4090 For low motion zero camera movement - LTX slaps and its soooooooooooo fast
10
63
389
25,835
This is so cool. This edit was done all inside of ComfyUI! The cuts were done based on a BPM analyzer and the videos were generated with Runway API nodes So so so clever
Lovis Odin
17
32
383
30,156
The goat Kijai back to cooking up magic This is VERY promising stuff: Video on left is source, video on right is completely generated locally on a 4090. Motion inference!
11
47
385
68,950
I see no scenario where I'd ever use Midjourney again Created with FLUX [DEV] running locally on a 4090 Upscaled 2x PROMPT IN ALT
29
32
376
33,483
👾 Another pixel art vid2vid I did this year using dead pool movie footage 👾 Process was my usual: 1) Img2img every frame with unsampled noise 2) clean it with AnimateDiff 3) DaVinci resolve Used Azib’s pixel checkpoints for XL and 1.5
A.I.Warper
13
35
381
28,424
🧵FULLY GENERATED AI ANIME THREAD I am not a "creative", but I will explain in this thread exactly how I made this video, fully AI generated, in less than an hour..... Imagine what a professional could do 😂 Please like and reshare if this helps or interests you 💙
18
46
391
29,297
Optimistic about this one 👀
Hyper3D by Deemos
8
37
369
35,571
The thing people may not realize is once you have this motion, you can remap any lip sync and eye movement to the video near perfectly with LivePortrait You just need the performance first Soon it will be so easy
Chubby♨️
25
29
352
44,218
Managed to get this much better and closer to "anime" You have to remember - this is a zero shot method. A few in the Discord have trained character loras and it's insanely good.... nearly perfect tbh
24
27
379
40,440
I’ve succeeded to make a meme template so viral, the meme lord himself has made one!! 🥹🥹🥹🥹🥹🥹🥹🥹🥹🥹
Elon Musk, when he posts a banger on 𝕏: nitter.app/_electronicCash/status…
24
12
365
34,806
Flux [DEV] locally w/ ComfyUI 1024 x 1024 gen upscaled 2x Spicyyyyyyyyyyyy details and the prompt adherence is insane. I never once mentioned the words Frieren OR Fern. (prompt in 1st comment below)
11
27
360
38,872
🐺The "Wolf" of Wall Street 📈 600+ frames on a 5090 Time: 35mins 20 seconds 1024 x 576 res and 4 steps - going to see if I can push the quality higher on a bigger card. Thoughts? The paws are my biggest critique. Believe it's because they exit frame a lot.
24
25
370
54,225
Bruh … 🦋 This is the ultimate slop factory….. I kinda dig it 🥲
6
32
362
25,430