Youtube comments of (@bycloudAI).

  1. we will soon live in a world where zack becomes a hot asian chick reacting to content and we will all think it’s real
    2300
  2. day 7482 of bycloud pronouncing something wrong: it's Touhou project (Japanese pronunciation) not DongFang Project (Chinese pronunciation) please don't revoke my weeb license I swear it was an accident
    1600
  3. This video is being picked up by the algorithm for some reason, so bear in mind that some information is a bit outdated even though it was only published like a month ago lol And if you want to learn more about the latest AI Research news, check out my latest upload ;) you will not be disappointed! https://youtu.be/rCygkyMuSQo
    680
  4. FINALLY I am only 1 day late to AI news cuz I spent the last 10 hours making this video so please like and sub ♡
    615
  5. Sup guys, since this video is popping off on the YouTube Algorithm If you do like this video, lemme shameless plug this AI voice cloning video that is also as fascinating as this AI which you may find interesting: https://youtu.be/MjaE0FjDHc8 Thanks for the support <3 It really means a lot to be able to grow while making fun and cool content!
    557
  6. Claude 2.1 has had some bad blood with needle in a haystack benchmark before so my hot take is that the people finetuned it added the "self-awareness" into Claude 3 as an easter egg when u test it lol
    431
  7. A more in-depth tutorial is on their project page I'll link that here: https://youtu.be/LQFkiBDgVIE However, some functions are not available for the web demo!
    401
  8. read abstract, follow discussions, read more in-depth if a paper is interesting To also speed up the reading process recently, I am also testing AI summaries to give myself a faster understanding probs will make my own AI summaries quality benchmark if I have the time too The benchmark will probably made out of how many key ideas the AI can address compared to my own take away from the papers The quality of summarization is still pretty low, or in some cases, a lot of prompt engineering has to be done to get LLMs to provide good summarizations
    383
  9. Apparently my dumb ass forgot to add the photo at 3:12 so if you are wondering how it looks like: https://imgur.com/a/HyJD10k thanks to MASTER0o for pointing it out lol
    383
  10. Please don't be weird in the comments 😭 and a new automated merging method just came out right when I finished this video... These technologies are moving too fast man you can check it out here: https://github.com/Xerxemi/sdweb-auto-MBW I might make a video about it if it really pops off, so subscribe to stay tuned!
    378
  11. Like this comment if you still don't have dalle 2 access edit: as of Sep 27th, I still don't have access. Am considering reapplying for the 3rd time LOL edit 2: the moment I made my first edit they opened access for dalle 2 nice. Everyone can use it now!
    252
  12. This is like seeing text to image synthesis a year ago, t2v is definitely going to the moon! Also, keep exploring at http://brilliant.org/bycloud/. Get started for free, and hurry—the first 200 people get 20% off an annual premium subscription. Oh and check out my latest video on Text To Video Generation!!! https://youtu.be/aux13apSEP0
    192
  13. 10/10 it's so cursed that I would doubt my life again
    155
  14. Oh ya, I'm currently looking for a video editor, don't need to be highly experienced but just need to have some knowledge in AI. Dm me on discord if you're interested! and check out https://brilliant.org/bycloud to start your interactive learning journey 😎😎 PS. Next video 100k Q&A (finally). Throw some questions at me down here! CORRECTION: 12:11 Waifu Diffusion 1.5 isn't the base model for any of the finetune and merge for anime on civitai WD 1.5 is based on SD 2.1 Most (All) anime SD 1.5 models are based the leaked NovelAI checkpoint thanks ash
    142
  15. for people wondering why i draw similarity with LK-99, it's because the results are not reproducible. there were too much ground to cover, so I skipped a lot of technical terms. let me know if you have any questions
    116
  16. Second time making a semi-documentary, do you enjoy these kinds of video style? Please let me know! And please stop flaming how I pronounced DALL-E I swear the extra E stands for my shitty English 😭 And of course, the first 1,000 people to use the link will get a 1 month free trial of Skillshare: https://skl.sh/bycloud06221
    111
  17. yall openai sales bots aint gonna get me in my own comment section
    81
  18. Get another browser that’s literally better at everything: https://operagx.gg/bycloud2 Sponsored by Opera GX! P.S. Looking for a video editor! DM me on Discord if you're interested
    75
  19. Samsung was changed into Samsong cuz my editor was making fun of my pronunciation. Why am I being abused again >:( And remember to visit https://brilliant.org/bycloud/ to get started learning STEM for free, and the first 200 people will get 20% off their annual premium subscription. oh and i made an installation tutorial for DaGAN to run locally: https://youtu.be/LmS22DoJNsA
    73
  20. Check out the full K/DA - Villian 60FPS here: https://youtu.be/YZy6l2dBbR4 Longer RIFE vs DAIN comparison compilation: https://youtu.be/2BwC_utLhH8 RIFE APP with Windows exe is available for free: https://nmkd.itch.io/flowframes
    73
  21. watch me release the next video next month LOL
    69
  22. To plug the sponsor: try everything Brilliant has to offer free for a full 30 days, visit https://brilliant.org/bycloud . The first 200 of you will get 20% off Brilliant’s annual premium subscription! P.S. Nothing in this video is voiced by a real person. All the voices are fake (except for 12:32 lol) The first 1 min (0:00~0:58) is generated using voice2voice with my real voice as the reference. 0:58~12:47 is generated with the combo which I mentioned in 11:46. From 11:46 till the end is all ElevenLabs Pro Voice Cloning.
    68
  23. The comment section is gonna be a shitfest, so brace yourself But I finally released this extremely delayed video, hope I clarified something for you in this one! And also check out NordVPN at https://www.nordvpn.com/bycloud if you are looking for a VPN and get started to browse the internet safely & privately ;)
    68
  24. I enjoyed too much making this video. It was a fun ride. The dataset was only around 400 wav files of 2~12 seconds audio too, which is crazy compared to how well this performed. The first 1,000 people to use this link will get a 1 month free trial of Skillshare: https://skl.sh/bycloud09212 check out my Goodbye Sengen cover here: https://youtu.be/5rEQfzds-WY or to ruin ur ears with harumachi clover here: https://youtu.be/Ht0-fqzrMHA
    66
  25. This video took too many revisions, hope you enjoy it, lmk if I missed anything too! PS: Break up with basic browsers! Get Opera GX here: https://operagx.gg/bycloud3
    65
  26. Here's a pretty interesting timeline that someone compiled, a pretty cool perspective one i would say The main takeway here: Sam & GDB non-stop raise funds for OpenAI after Dev Day cuz of its success, planning to have a 3 times evaluation of OpenAI from 30 Billion to 90 Billion, and looking to fund OpenAI even further. Ilya was unsettled by Sam's fundraising moves, might have wanted it to stop before they get in trouble. The lack of comms might have been fundraising without board permission from this perspective. https://x.com/8teAPi/status/1725724907722752008?s=20
    64
  27. had to remap the project files, probs screwed up there oops But maybe DeepSeek is just Meta but for Chinese, so close enough :D
    62
  28. let me know if you guys want to see the entire lore of the hottest open source LLM debate on twitter by Yann LeCun last month
    61
  29. oops i meant it metaphorically that was a bad word choice lol
    60
  30. The artist community is probably going to go insane about this line art colorization tool now, hopefully it'll be positive D: ad: Check out Opencv's New AI Art Course Kickstarter at https://bit.ly/410U3Xs and join their AI Art Contest now to win an iPad Air! https://form.jotform.com/230370613300439 And check out my Patreon if you wanna support my work!! https://www.patreon.com/bycloud
    59
  31. FYI, Gura voice clone was trained with around 40 mins of her audio. Special thanks to Justin John for assisting me too. You can learn more about voice cloning AI here: https://youtu.be/MjaE0FjDHc8 or even get NordVPN here: https://nordvpn.com/bycloud or use my coupon code "bycloud" to support me!
    58
  32. This is literally gold, you have explained everything so intuitively and made it so much easier to understand! Thank you so much Andrej for sharing this in-depth knowledge for free!
    55
  33. check out HubSpot's Free AI Task Delegation Playbook here! https://clickhubspot.com/s6ef As for the new paper gamegen-o I've shown at 12:00, not actual interactive demo is shown. Looks like the demo is heavily cherry-picked. So yes, gamengen is by far the closest example to an actual AI game engine. edit: Looks like they deleted their gamegen-o project, with no reason specified. Odd for sure.
    54
  34. I am only 38 days late covering controlnet 1.1, not too bad ig oh and the first 1,000 people to use the link will get a 1 month free trial of Skillshare https://skl.sh/bycloud05231
    54
  35. Try Mammouth now for only $10/mo! https://mammouth.ai With the release of the full o1 and o3 model, this video has unfortunately already been outdated! (I am too slow as usual oops) But these are still some really good models which you should keep your eyes on, cuz they are gonna be the ones you are able to use locally in the future ;) I am also making a summary on o3, o1 full/12 days of OpenAI, so stay tuned!
    52
  36. 4hrs while being 4k quality is chef kiss
    51
  37. real
    49
  38. Merry Christmas yall! Here's a fun video for everyone before 2023 ends:)
    48
  39.  @sshah5800  twitter unfortunately
    48
  40. I need to stop stealing "a few more papers down the line" but that line is just goated
    48
  41. Check out ThinkBuddy using the code "BYCLOUD" in the link: https://thinkbuddy.ai/ltd to get your discount!
    48
  42. If you want to train, finetune or run models that needs high VRAM or computational costs, check out today's sponsor RunPod to get started on the best priced GPU cloud services using this link: https://l.linklyhq.com/l/1jHv6
    47
  43. if only I am also called Riley Goodside, maybe I can bank in 300k a year
    46
  44. ya unfortunately my videos take longer to make, since I want the editing in top notch quality. And OpenAI is the one that actually discussed about MineRL which is a competition I've been following for quite a long time, so I really wanted to at least cover that lol but ya you're right MineDojo is the chad here. Sorry I ain't the latest AI news content creator, feel free to turn to Yannic or anyone else that literally have ML news discussions weekly!
    45
  45. I was a tiny bit suspicious when xLSTM didn't publish codes, but damn okay
    44
  46. I am actually planning something similar like this, it'll be sick
    44
  47. I think if you press the "join" button beside "subscribe" you will instantly understand the video
    43
  48. i need to get off twitter and stay on arxiv
    43
  49. Made a crappy installation tutorial, if you have any questions ask me on my discord channel https://youtu.be/ICvNnrWKHmc
    42
  50. Oh ya I am wrong again. It's "Pie"fuHD not "Pee"fuHD please don't cancel me, and yes ICON supports .OBJ, so it's very convenient. The Chun-Li kick result was semi-clickbait oops: https://imgur.com/a/pw0Ry5j oh and also please visit https://brilliant.org/bycloud/ so daddy brilliant will sponsor me again cuz I love their courses and they are genuinely making some amazing content aka: visit https://brilliant.org/bycloud/ to get started learning STEM for free, and the first 200 people will get 20% off their annual premium subscription!!
    41
  51. This video is in 4K, view in 4K for the best quality. And feel free to comment down what you think! Remember to check out Yours App using this link: https://yoursapp.com/bycloud or use the code: BYCLOUD to save 60% off the yearly plan!
    40
  52. Another aspect is that we are reaching a point where there will be infinitely more content than we can humanly consume.
    40
  53. What do you guys think about my new video format? Let me know down here! and the first 1,000 people to use the link or my code bycloud will get a 1 month free trial of Skillshare: https://skl.sh/bycloud05221
    40
  54. There is currently a huge controversy over this fully AI generated south park research, some are panicking and some says it's a hoax. South Park makers already have a track record for making an episode scripted by ChatGPT, so I guess we'll find out what's real and what's not next week... Oh and check out https://Tensor.Art/ now to start generating images online for free! I also had a mix up at 10:22, it's supposed to show Bard not Bing. I was gonna cover both but somehow forgot and mixed them together lol
    39
  55. Try Mammouth now for only $10/mo! https://mammouth.ai also let me know what you think about the new style of explanation down here!
    39
  56. /s
    38
  57. There was quite a long delay on AI research videos, but I am back! The video flood gate has just opened, quite a few videos will be out soon too so stay tuned! And keep exploring at http://brilliant.org/bycloud/ . Get started for free, and the first 200 people get 20% off an annual premium subscription! P.S. also looking for video editors interested in AI related stuff, dm me on Discord if interested;) and also, AFL = animefull-latest
    36
  58. that's rough buddy
    36
  59. Let me know if you guys want a dive into the methodologies of TTC, there's a lot of new papers/implementations coming out every day lol (entropix is a cool one) Check out NVIDIA's suite of Training and Certification here: [NVIDIA Certification] https://nvda.ws/3XxkFyj [AI Learning Essential] https://nvda.ws/4gvD474 [Gen AI/LLM Learning Path] https://nvda.ws/4enwYE7 You can use the code “BYCLOUD” at checkout for 10% off!
    36
  60. Check out HubSpot's ChatGPT at work bundle! https://clickhubspot.com/twc and I missed out a paper that combined mamba & attention, you can find it here: https://arxiv.org/abs/2402.04248 what an interesting timeline this is
    33
  61. oh i follow https://mail.bycloud.ai/ it highlights some of the latest best papers
    32
  62. maybe i'll make a video on it (and it'll instantly get outdated
    32
  63. going for the "ranking by views: 10 of 10" for this one 🔥🔥🔥🗣️🗣️🗣️
    31
  64. Fun fact, the guy in a top hat (1:27) is sent in from viewer requested to be upscaled and restored. I've kept it for over 6+ months as a baseline to test the face upscaling functionalities, and GPEN is the only one that I had success in! (pls don't email me restoration requests tho, the person got lucky cuz I took great interest myself too)
    30
  65. In fact it can. DALL-E is basically like a different way of generating, it definitely needs its own video. I'm planning to make a video about AI generated art soon, and I'll cover it.
    29
  66. Try Mammouth now for only $10/mo! https://mammouth.ai
    29
  67. check out my main explainer video about this voice cloning AI: https://youtu.be/MjaE0FjDHc8 or to ruin ur ears with harumachi clover here: https://youtu.be/Ht0-fqzrMHA
    29
  68. google is not really transparent with its specs, so that, I dont know either tbh
    29
  69. If you gonna call me out cuz i cited wikipedia, imma accuse u for being my middle school history teacher And keep exploring at http://brilliant.org/bycloud/ Get started for free, and hurry, the first 200 people get 20% off an annual premium subscription!
    29
  70. ngl I didn't expect OpenAI's "strawberry" to actually be this cool oh and u can join my newsletter here! https://mail.bycloud.ai correction: they are still gonna continue the GPT series, so they now have 2 different model series in development
    28
  71. maybe we all are just little minions to collect data for AIs
    28
  72. To try everything Brilliant has to offer—free—for a full 30 days, visit https://brilliant.org/bycloud . You’ll also get 20% off an annual premium subscription! Flux.1 is also now available on X with Grok-2, but I think you need to have subscription to use it
    28
  73. A quick clarification: 3:33 to 3:58 of the cooking and the day in a life are all teleoperated, not behavior cloned. It was meant to be a demonstration on what teleoperation can do, and what behavioral cloning can potentially do with these teleoperated data.
    28
  74. Start a VPS cost effectively with Hostinger today and use discount code "BYCLOUD" for an extra 10% off! http://hostinger.com/bycloud As for DeepSeek's MOAT, it's pretty much their talent: "They allegedly offer salaries of over $1.3 million dollars USD for promising candidates" via SemiAnalysis https://semianalysis.com/2025/01/31/deepseek-debates/ update: rumors said that they might only have 2000 H800s, 10k A100s (confirmed), some H20s, and they dont even have up to 50k GPUs. Will update soon
    27
  75. Check out NVIDIA NIM now using this link https://nvda.ws/3Jn5pxb It seems like the general consensus that people are disappointed with Stable Diffusion 3 is because there are so much potential wasted. I definitely think it's a bit unfortunate too... EDIT: Update on new SD3 terms The license was changed and is now free for research and also commercial use up to $1m. - Lykon Also some elaborations by Lykon (key researcher for SD3) 1. 2B was an experiment at scaling down 8B, but it's essentially the same architecture. As Alex said on Discord, it was supposed to be released as "beta", but the label got removed last minute. That was a mistake. 2. 4B is very experimental and, as of today, has roughly the same issues as 2B. It's also using a different architecture and only one text encoder (which is probably the worst of the 3 and the heaviest). 3. 2B was released before 8B because it's easier/cheaper for us to finetune, and is also the perfect size for the community (having roughly the same resource cost of SDXL). 4. The new CEO just changed the license entirely making it free for research purpose, free for commercial use until 1M USD, free for non-commercial use. 5. We are working on a new Medium model to address the issue. 6. The issues with 2b were not intentional. The architecture was made and tested at 8B params and, turns out, MMDiT scales very well at high param count, but has attention issues on low param count. We are addressing this. 7. The 512mar SD3M that Alex showed is not the finetuned one. I just re-tested the one I have today on "that prompt" and it works much better. I'm pushing to release it (even if it's not super useful)
    27
  76. Sign up Shipd now to start earning while coding! https://tally.so/r/3jBo1Q And check out Datacurve.ai if you're interested: https://datacurve.ai/ On a side note, I am also looking for some like-minded people that are down to work together. For video scripting to maybe revive the AI newsletter with me, feel free to hit me up on Discord if you’re interested!
    27
  77. Sign-up for GTC24 now using this link! https://nvda.ws/48s4tmc RTX4080 Super Giveaway participation link: https://forms.gle/2w5fQoMjjNfXSRqf7 oh and check out my AI website leaderboard if you're interested! https://leaderboard.bycloud.ai/ Here's a few corrections/clarifications for the video: - it's named 8x7B but not 56B is because it's duplicating around 5B of Mistral-7B's parameters 8 times (=40B), then there's the extra 7B for attention and other stuff (40B+7B=47B) - experts are not individual models, but only FFNs from Mistral-7B - experts operates not just on a per token level, but also per layer level. Between each feed-forward layer, the router assigns that token to an expert, but each layer can have a different expert assignment. - Router is layer based, and it decides which expert for that next layer to activate, thus it looks different over each layers 3:26 and you can refer to this diagram https://imgur.com/a/Psl1Fi4 an example process of activations that might happen for any given token, and it's picking a new set of 2 experts for every layer special thanks to @ldjconfirmed on X/twitter for the feedbacks
    26
  78. when i put in csgo literally no one gave a fuck maybe valorant really is overrated
    26
  79. An official ANIME MODEL for upscaling is out! Refer to their official repo here: https://github.com/xinntao/Real-ESRGAN/blob/master/docs/anime_model.md The first 1,000 people to use this link will get a 1 month free trial of Skillshare: https://skl.sh/bycloud08212 Start learning some nice & creative stuff today!
    25
  80. got it, thanks for the feedback guys:))
    25
  81. The first 1,000 people to use the link will get a 1 month free trial of Skillshare https://skl.sh/bycloud07231 ! and please keep in mind that this video is to raise the awareness of C2PA + how dumb it is, and not trying to address, attack or criticize artists like Karla Ortiz on them advocating for artists rights.
    25
  82. oh ya I'm looking for an editor, dm me on discord if u are interested
    24
  83. This video might be a bit late, but hopefully it provides a good enough introduction for you to get into generating AI art! Check out OpenArt's online AI ART Generator and Photobooth now at: https://openart.ai/
    24
  84. nerf vs 3d gaussian splatting is gonna be an interesting research topic 1v1 for the next year lol and to try everything Brilliant has to offer—free—for a full 30 days, visit https://brilliant.org/bycloud . The first 200 of you will get 20% off Brilliant’s annual premium subscription!
    23
  85. bro u havent even watched it 😭
    22
  86. I would tend to agree too, I would say it's more of a style transfer/converter than looking like an anime character. I tried to express this in my description but I guess I probably should've mentioned this part in the video.
    22
  87. This video's background music are all generated by AI, what do you think? oh and to try everything Brilliant has to offer free for a full 30 days, visit https://brilliant.org/bycloud The first 200 of you will get 20% off Brilliant’s annual premium subscription!
    21
  88. just gotta gaslight harder at this point lol
    20
  89. My sponsor actually told me "strange sounds of chatter and laughter in the first 20 seconds of the video", was my opening skit actually that bad...? But ya, tired of getting "CUDA out of memory" on colab? Give Q Blocks' RTX3090 a shot: https://www.qblocks.cloud/invite/bycloud
    20
  90. Or even 3D scene. You can have a whole ass room or even a stadium like their official demo. It's mind-blowing
    20
  91. my mouth actually hurts after recording
    18
  92. To try everything Brilliant has to offer—free—for a full 30 days, visit https://brilliant.org/bycloud/ . You’ll also get 20% off an annual premium subscription! (I reuploaded this video cuz there was a pretty big mistake at 6:13, sorry notifications!)
    18
  93. there's actually a new paper about using both RAG and context window, by routing queries to RAG or long context depending on self-reflection, p cool u can check it out https://arxiv.org/abs/2407.16833
    18
  94.  @dscodes_  I think they follow https://x.com/TheAITimeline on twitter for weekly top paper recaps!
    18
  95. Check out Imagine.Art at https://www.imagine.art/ oh and also, SDXL just announced it'll be released OFFICIALLY on July 18th! https://twitter.com/EMostaque/status/1676923097159712768?s=20
    17
  96. tbh just a matter of time to decrease, and it can already run on 12gb which is not that bad compared to the amount of stuff it runs
    17
  97. BROTHERS and SISTERS HERE"S THE FREE non-technical guide for using AI in your business: https://clickhubspot.com/fuu7
    17
  98. stay up-to-date on the latest AI research with my newsletter! → https://mail.bycloud.ai/ Minor correction: GGUF is not the predecessor to GGML, GGUF is the successor to GGML. (thanks to danielmadstv)
    16
  99. You can try out Luma AI's Dream Machine here! https://luma.1stcollab.com/bycloudai I am really good at having great timing. MovieGen came out when I nearly finished the video. I'm sad. So here's a quick definition of DiT: A diffusion transformer (DiT) is a model that combines elements of diffusion models and transformers to generate data like image synthesis, audio generation, or text generation. Diffusion models are a class of probabilistic generative models that create data by iteratively denoising a latent variable, which starts from pure noise and is gradually transformed into a coherent sample. Transformers on the other hand, are neural network architectures known for their ability to model long-range dependencies in data, primarily through self-attention mechanisms. You could ultimately say that, a diffusion transformer is just a transformer with the goal of denoising. Yum. Here's MovieGen's paper: https://arxiv.org/abs/2410.13720 it contains a better run down to crafting the latest near SoTA video generation
    16
  100. I think the colab is pretty straight forward to use itself, but if you need help, feel free to ask on my discord! Edit: here's a quick rundown on how to run the colab: https://imgur.com/a/BPiHxDk And the first 1,000 people to use this link will get a 1 month free trial of Skillshare: https://skl.sh/bycloud01221 Check it out if you want to learn anything!
    16
  101. hmm interesting, i’ll check with the pytorch implementation if that’ll be fixed, just windows things ig would u mind sharing ur results too?
    16
  102. The first 1,000 people to use this link will get a 1 month free trial of Skillshare: https://skl.sh/bycloud09211 and how's my new mic?
    16
  103. To try everything Brilliant has to offer—free—for a full 30 days, visit https://brilliant.org/bycloud/ . You’ll also get 20% off an annual premium subscription.
    16
  104. Check out HubSpot's Free ChatGPT resource to power up your work efficiency🔥https://clickhubspot.com/hyx
    15
  105. Check out D-ID now and start synthesizing your own avatar with the free 5 minutes credit! https://timf.io/3Tv5Gob Also there are 2 new research released called Fairy video2video https://fairy-video2video.github.io/ and PIA https://pi-animator.github.io/ which I missed so I'll link it here lol, the arms race is still going?!
    15
  106. Scale your business easily now with AI/ML API to deploy any models you need! https://aimlapi.com/?ref=bycloud and Claude 3.5 Opus is about to be glorious
    15
  107. What do yall think about the explanations or should I skip more next time? If you like to learn more yourself though, the first 1,000 people to use this link will get a 1 month free trial of Skillshare: https://skl.sh/bycloud03221
    15
  108. LOL
    15
  109. I will try better next time 😭
    15
  110. Check out HubSpot's Free 1000+ Prompt resource here! https://clickhubspot.com/awmh POV: even sam altman denied AGI is not here, and not here next month either https://x.com/sama/status/1881258443669172470 (so it's definitely march)
    15
  111.  @tf5pZ9H5vcAdBp  gaslight the AI to do things you want it to do
    14
  112. Secure your business effortlessly with a limited time offer of 3-month NordPass trial! Use ”bycloud” activation code at nordpass.com/bycloud I felt like I had to share the stuff in this video for quite some time, so it's a different type of video than usual! Let me know what you think here P.S. for number 5 im snitching on yall to your parents
    14
  113.  @Serizon_  u lowkey missed out bro
    14
  114. Go to https://strms.net/babbel_bycloudai to get up to 60% off! #ad and I am also thinking to do a 100K Q&A, feel free to ask me questions here or on my Discord's special Q&A channel! https://discord.gg/NhJZGtH
    14
  115. ESRGAN 0.5FPS Real-ESRGAN 0.6FPS Real-CUGAN 4.5FPS Waifu2x 3FPS Topaz 2FPS All ran on a RTX3090 with VSGAN TensorRT Docker Didn't put it as a criteria because it's too easy to show what's good and what's not
    14
  116. Might be a bit late about the llama-3.2 news as usual oops. Working on the New Claude 3.5 Sonnet video now (see u in 3 weeks lol and check out NVIDIA's Llama-3.2 resources 👇 Jetson AI Lab: https://nvda.ws/4eO2VFU Multimodal RAG with Llama 3.2 GitHub: https://nvda.ws/4eyspY0 NVIDIA NIM for Developers: https://nvda.ws/4dCXys1
    14
  117. loved your reactions, can't believe I live in a timeline where Asmongold is watching my video
    14
  118.  @theodorenevermore6672  i mentioned in the main video that I used a reference vocal to get the right pitch. So ya, some parts I have to sing it physically to get the right pitch.
    14
  119. Check out HubSpot's Free ChatGPT resource! https://clickhubspot.com/bycloud-chatgpt and as usual, I am the Golden Gate Bridge 😎https://mail.bycloud.ai/
    13
  120. Check out HubSpot's Free 1000+ Prompt resource here! https://clickhubspot.com/kiof
    13
  121. Join the waitlist now for exclusive access to the OnDemand platform: https://on-demand.io/contact let me know if you like these kinds of research breakdown too! my newsletter: https://mail.bycloud.ai/
    13
  122. oh ya I am planning to make a 100K Q&A, feel free to ask questions here or on my discord! https://discord.gg/NhJZGtH thanks for all the love and support guys <3
    13
  123. check out Storyteller AI at www.storyteller.ai/?ref=bycloud and join their discord now! https://discord.gg/storyteller this new age of internet is about to be chaotic 💀
    13
  124. Check out Monica AI now, the best AI browsing companion! http://bit.ly/3Qh6zxy and use the code BYCLOUD10 for a special 10% off PS It is a really nitpicky video, a lot of capabilities is really limited to the model's performance, and not the information retrieval process. Currently on all other models that I tested, no model can properly explain DeepSeek's "cold-start" data yet. So I just wanna say again that Deep Research is amazing, especially with it capable of writing in long form. It's just that it might be on the model's part for a lot of its downfall, especially Deep Research's o3 is RL tuned specifically, which worsen its normal o3 capabilities.
    13
  125. I do not own the latest version of photoshop, so it's kind of unfortunate that I cannot do a side by side comparison:/ You can try out the AI here on this website: https://cleanup.pictures/ or you can get NordVPN here: https://nordvpn.com/bycloud or use my coupon code: bycloud
    12
  126. Don't miss out on these exciting upgrades designed to elevate your content creation experience with DomoAI! Go try out: discord.gg/sPEqFUTn7n
    12
  127. Take your personal data back with Incogni! Use code bycloud at the link below and get 60% off an annual plan: http://incogni.com/bycloud maybe we are all bots and the dead internet theory is true
    12
  128. Automatic1111 has a SDXL PR now! https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/11757 The x.AI Twitter Space is happening 1PM PT Time https://twitter.com/xai/status/1679527914394763264?s=20 And dont forget to check out ImageCreator, the free AI plug-in for Photoshop now using this link: https://imagecreator.alkaidvision.com/ !
    12
  129. Try out Poe now and save your $$ on multi-subscriptions! https://quora.1stcollab.com/bycloudai and probs no more 20 mins vid from me it's literally death itself to record it
    12
  130. So far no, but hopefully it'll be available soon.
    12
  131. Check out HubSpot's ChatGPT at work bundle here: https://clickhubspot.com/2os unfortunately topping the last mamba edit is way too hard, but I guess now at least we know mamba is real
    11
  132. To try everything Brilliant has to offer—free—for a full 30 days, visit https://brilliant.org/bycloud . The first 200 of you will get 20% off Brilliant’s annual premium subscription! Making this video was quite a rollercoaster! From Dall-e 3 not yet been releaed, to confirmed multi-modal GPT-4 release, I cannot believe I have hijacked such a funny timing.
    11
  133. To try everything Brilliant has to offer—free—for a full 30 days, visit https://brilliant.org/bycloud/ . You’ll also get 20% off an annual premium subscription! Like this comment if you wanna see more MoE related content, I have quite a good list for a video;)
    11
  134. The first 1,000 people to use the link will get a 1 month free trial of Skillshare https://skl.sh/bycloud08231 and here is the list of all the best tools I mentioned in this video Paid options: Colorization - https://www.Palette.fm Damage Restoration -https://www.MyHeritage.com Facial Upscaling - https://www.MyHeritage.com Free options: Colorization - DeOldify https://deepai.org/machine-learning-model/colorizer Damage Restoration - Bringing Old Photos Back To Life https://replicate.com/microsoft/bringing-old-photos-back-to-life Facial Upscaling - Codeformer https://replicate.com/sczhou/codeformer
    11
  135. Whether you want to use Flow irl or implement in your business, check it out here: https://www.speechmatics.com/flow?utm_source=bycloudai&utm_medium=paid-media&utm_campaign=flow-2025&utm_content=youtube
    11
  136. First time recording while not high. Not bad. for legal reasons, it was a joke. I just recorded in the afternoon when my voice is not fucked lol
    11
  137. A different video than the usual, what do you think? And thanks again to AI Camp for sponsoring today's video! Check out their amazing AI summer camp now: https://ai-camp.org/partner/bycloud
    10
  138. Sign up for NVIDIA GTC2025 here: https://nvda.ws/48s4tmc And Join The RTX4080 SUPER Giveaway (enter between March 17-21st) https://forms.gle/TbGgoD5obn1zRz5r7
    10
  139. To try everything Brilliant has to offer—free—for a full 30 days, visit https://brilliant.org/bycloud . The first 200 of you will get 20% off Brilliant’s annual premium subscription! Shoutout to DAN and rip DAN 2023-2023 🙏
    10
  140. To try everything Brilliant has to offer—free—for a full 30 days, visit https://brilliant.org/bycloud/. You’ll also get 20% off an annual premium subscription. I have a few more interesting ways of how people use model distillation, lmk if yall would like to see more creative use of this idea!
    10
  141. ya they sent me on discord
    10
  142. you the goat, thank you so much for your kind words!
    10
  143. Check out Gamma.app now using this link here! https://gamma.1stcollab.com/bycloud
    10
  144. The first 1,000 people to use this link will get a 1 month free trial of Skillshare: https://skl.sh/bycloud07212 Have fun learning well organized courses for 1 month free!
    10
  145. Want to compare the results from the chatbots yourself? Check out ChatHub now! They offer one of the easiest ways for you to compare and evaluate the results to find the best LLM for you! https://chathub.gg/?utm_source=bycloud
    10
  146. oof
    9
  147. well it's technically my fault cuz i screwed around with the nested sequences to fix something, it might just be my nemesis but i guess new viewers will probs not notice the second one so for every viewer that notices pls like this comment and as a punishment for myself i'll do 1 push up for every like if u just like this comment for fun i'll not wish u a happy halloween
    9
  148. The first 500 people to use my link will get a 1 month free trial of Skillshare! https://skl.sh/bycloud10241
    9
  149. Streamline AI task delegation with HubSpot's Free Playbook: https://clickhubspot.com/9yu and check out my newsletter 😎 https://mail.bycloud.ai/
    9
  150. Check out HubSpot's FREE AI Prompt Library Now! https://clickhubspot.com/h14h
    9
  151. actually, I found a very recent paper that is way better than this and imma use that in a future vid soon to demonstrate it to you all.
    9
  152. Facebook AI (aka Meta AI) research lab is actually expert in object detection + object segmentation. Research like Detectron2 https://github.com/facebookresearch/detectron2 would also be a key. aka it slipped my mind that facebook research is actually good at detecting objects
    9
  153. The first 1,000 people to use this link will get a 1 month free trial of Skillshare: https://skl.sh/bycloud07211 Check it out if you want some pretty pog AI/ML course Codes of Alias-Free GAN are released under the name StyleGAN3!! Check it out here: https://github.com/NVlabs/stylegan3
    8
  154. Check out OnDemand now! https://app.on-demand.io/auth/signup?refCode=bycloud and please give back my copium canister
    8
  155. Check out today's sponsor InWorld AI with the link here: https://bit.ly/inworldbycloud and start crafting an even richer world setting for your game! also, I am looking for video editors, dm me if you're interested
    8
  156. it's like RNN's hidden states are just ML models thanks for watching till half way tho <3 time to go back to drawing board
    8
  157. yeh, it was just a way for me to show people's skepticism and nothing else
    8
  158. ​ @cvs2fan  their business model seems to be betting on an user's usage doesn't exceed a certain amount monthly since to reach 10$ a month is still pretty hard to hit for basic QA usage unless you are using the models heavily for things like coding where it uses a lot of context and even if you do, they might be betting on the months you dont use it as much these are just my guesses though idk if they have any hard usage limits, but I think so far mammouth is the only all-in-one that i use too and the quality seems to be consistent to what they advertise and i use a lot of mammouth's image gen models since i get mj recraft and flux 1.1 without having the need to go to their websites and pay an expensive subscription just for 1 or 2 images
    8
  159. sorry guys, missed out a segment while rendering so I had to reupload!
    7
  160. @@thelegendguyofficial dw the music and the content is not HAHAHA and will probably not be anytime soon here's the music yt link https://youtu.be/0RygB-mapsk this person makes banger lofi, go support them
    7
  161. the old one is really inaccurate. I should have made a comparison shouldn't I :/
    7
  162. Seems like I wasn't clear enough in my video. They might have slightly better or worse models to estimate 3D positions but at the end of the day it relies on monocular 2D estimations first, then convert into 3D. The same goes for the commercial options too. Essentially, they both relies on 2D estimations, it's just that the visualization for DeepMotion is able to be presented in 3D. That's all. A comparison video is literally about testing each of them in different use cases. This video is just to demonstrate to people the difference and similarity so that everyone know what they will get out of them when they use these technologies.
    7
  163. f
    7
  164.  @return_null  they are all up on my discord
    7
  165. damn im washed
    7
  166. you have some really interesting ideas!
    7
  167. RealESRGAN_x4plus_anime_6B ESRGAN model it's not Real-ESRGAN, it's pretty confusing naming method
    6
  168. oops ya that was my visualization mistake, should've been https://imgur.com/a/dIoEa0i
    6
  169. Start building your ideal daily routine! The first 100 people who click on my link will get 25% OFF Fabulous Premium: https://mtchm.de/6wi6m
    6
  170. Happy FlexiSpot Black Friday Sale now, Up to 65% OFF! You also have the chance to win free orders during this period. Use my code ''YTE7P50 '' to get EXTRA $50 off on the FlexiSpot E7 Plus standing desk USA: https://bit.ly/3OiLQJ5 CAN: https://bit.ly/3ZiPE3a
    6
  171. I put the characters in hard because they generated not as good lol but at least im glad u can guess it which means it is possible
    6
  172. will change it up
    6
  173. it's the length of the prompt plus the maximum number of tokens in the completion
    6
  174. thanks for your kinds words and ya I think I didn't present it clear enough at 9:19 I was referring to the lalal vocal isolation being less muffled compared to Spleeter. While it's the opposite as Spleeter has better vocal removal in that segment than lalal. To simplify what I said in my video: Vocal Isolation in that song Lalal > Spleeter (because lalal conserved higher frequencies) Vocal Removal in that song (Instrumental) Lalal = Spleeter (both has up and downs)
    6
  175. Hey Olivio, I definitely should have messaged you and chat with you about this earlier, idk why that didn't cross my mind I am sorry for like what I seem to be misinterpreting or misrepresenting what you have said, I just wanted to spread some more balanced views & information on it. Since I've lurked around a lot of discord/reddit mentioning your interpretation of the license, and there seems to be quite a lot of misunderstanding coming from the people that watched your videos. I've also consulted a few people on the terms that you mentioned in the video, and do think there's a bit of a misunderstanding, at least from what we understood from your video. Sorry for any misunderstandings I've caused.
    6
  176. It does not necessary mean it's not an AI research when it doesn't use neural networks, if that's what you are implying
    6
  177. and we dont even know if its real or not lmao
    6
  178. it's from Andrew Ng's lecture notes page 16, and taken out of context (my bad lol) you can find the PDF here https://stanford.edu/class/cs294a/sparseAutoencoder.pdf the notations usually shouldn't have numbers so it looked a bit confusing
    6
  179. thanks
    6
  180. F
    6
  181. I have an upcoming video about Mamba that is in the works, so stay tuned:))
    6
  182. i’ll pay for whatever shipping it costs unless the country is unshippable like north korea
    6
  183. bro u didnt even finish the video but thanks <3
    5
  184. thank you!
    5
  185. stylegan3 had solved the texture sticking problem. Must be how they embedded using stylegan2 instead of stylegan3
    5
  186. he's my goat
    5
  187. I literally played it on my screen and did it like that lmao
    5
  188. I think this creates a possibility to bring the overall standard of animation up, which at the same time improves our entertainment and let artists to focus more on refining details rather than in the need of spending time to color or paint. To an extent jobs come and go, but at the same time it'll create new job opportunities too, so don't worry!
    5
  189.  @davetwo7879  There are just too much AI news, so I decided to do these AI news videos lol cuz its kinda fun dw normal higher quality videos are still being made, but now I can just take more time to edit so I don't have to worry about I have nothing to upload
    5
  190. There are some vids on youtube that teaches you how to use it on the colab. https://youtu.be/xVQHgF4XpNM
    5
  191. yikes I think my TTS didn't grew up in NA
    5
  192.  @liuzeyu  i struggle to pronounce stuff fast, so for a 10 mins vid, it takes around 4 hours to record so for this 20 mins vid, it took like 8+ hours cuz my vocal cord gets sore from all the retakes to ensure everything sounds "okay" pronounced... and i get cold or allergy very easily, and that might also be a variable to slow down the recordings and creates even more retakes lmao
    5
  193. just be transparent, I thought the sponsor would be a great fit as they aren't selling anything for yall, and do provide some decent value to newcomers. if you still don't like it, sure i dont mind. But imho there are other way worse "sponsors", gambling platforms, fake couseling platforms, and this free one is worth the slander? Sure? I mean of course, no one likes to look at ads, but in exchange I might not be able to do this full-time and have people to help me with it. Well, it might not even be your concerns, as i'm just every other AI youtubers making videos, why should u care right? And when sponsor looks at all these negative comments about them, I wouldn't be surprised they will drop me, but that's not a problem for yall anyways. I don't delete any comments either, even if I dont respect your opinions or my feelings are hurt or at the cost of losing sponsors, I still leave it for the sake of transparency. Unless it's just straight up illegal or botting. (duh) As for the sponsor segment disclosures, I've done literally what every other youtubers done. For example in pewdiepie's latest video https://youtu.be/rpQswht3dsc he did the disclosures just like how I've done it, the only thing that differs is the product. I'd admit maybe the segway is a bit too seemless, but I just in general appreciate a smooth segway. Sorry if that comes off as slimy, maybe I'll adjust it in the future.
    5
  194. In the indie game devs point, I think I may not have explained it well. What I am trying to say is, we are at a point where the trade-off between convenience and sacrifice of jobs is hard to balance. On one hand, we can help potentially a few artists not lose their financial income, on the other, other people would then require to spend time & money in their life figuring out all those points you've just listed above, while an existing option is available BUT labelled as "unethical". Since the line that everyone draw their ethical standard is obviously different, I guess balancing what is "fair" between artists & everyone else is probably going to be the solution to settle things.
    4
  195. the same reasoning from the video applies
    4
  196. bycat voice reveal
    4
  197.  @ghulammahboobahmadsiddique8272  6:13 i showed class 3 chaos twice, i needa catch some sleep lol
    4
  198. oh LOL thanks for pointing that out not an ai, just terrible at speaking english
    4
  199. NFTs can sit
    4
  200. hey thanks for the clarification! just wanted to add, the reason why i attributed hunyuan is by hunyuan, is because on their paper, the team behind it are cited as "Hunyuan Foundation Model Team" and not for example Tencent AI Lab. But otherwise, you are right, I am pretty sure they are funded by Tencent.
    4
  201. I made one, this link will take u there https://youtu.be/_hcYgA-UjCc
    4
  202. I just clarified what was being misunderstood in the community about the licensing, and other than that, I said the literally same thing as you. Did you even bother to watch the video...?
    4
  203. 1v1 me yandere deepfake lul
    4
  204. vtubers but an actual face im not ready for the consequences
    4
  205. too fast too quick
    4
  206. llama3 herd of models paper
    4
  207. was worrying if it was too overwhelming, but thank you for your kind words:)
    4
  208. I wanted to talk about GRPO but that formula might be too scary for this video lol and i think for R1 training they used deepseek-v3-base (which they explicitly specified in the R1 report), not the SFT'd deepseek-v3 On page 23 (thanks for referencing btw), they were talking about applying SFT on the base model "Following this, we conduct post-training, including Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) on the base model of DeepSeek-V3, to align it with human preferences and further unlock its potential" so ya deepseek-v3-base is not a sft model but deepseek-v3 is. and r1 used deepseek-v3-base
    4
  209. DeepSeek-R1 paper page 3 https://arxiv.org/pdf/2501.12948 https://imgur.com/a/UvMhzbr
    4
  210. they only implied lighter shading = approach similar to best-of-N otherwise they are still using the thinking model
    3
  211. I love your narrative, what a great video! Definitely would love to see more!
    3
  212. it’s harder but not impossible. The main selling point for them is UX, chatgpt can’t accommodate those as well rn
    3
  213. u took an L
    3
  214. i think you are talking about image segmentation, see my other comment or just try searching detectron2
    3
  215. ikr terms keep evolving its crazy never heard of landmark spoofing actually and sounds pretty cool, I shall steal that in the future lol
    3
  216. ya was thinking of doing that, will add when i have time, thanks for the suggestion
    3
  217. oh shit
    3
  218. ​ @spicym4ngo  https://www.cambridge.org/core/services/aop-cambridge-core/content/view/0ACB73CB66134BFCA8C1D55D20BE6392/S1351324916000243a.pdf/the-return-of-the-chatbots.pdf https://en.wikipedia.org/wiki/ELIZA_effect#Significance https://en.wikipedia.org/wiki/History_of_artificial_intelligence#Natural_language https://web.archive.org/web/20110425191843/http://www.stanford.edu/group/SHR/4-2/text/dialogues.html
    3
  219. have no purpose just like my website https://leaderboard.bycloud.ai/ check it out 😎
    3
  220. thanks, your words have enlightened me
    3
  221. well tbh it started on twitter, i guess it wasnt a surprise that people made a big deal over it lol
    3
  222.  @minecraftmaniac2003  generate an image of a natural looking dark hair Asian girl with taking selfie in front of a mirror in a toilet holding an iphone 14 Pro in one hand and holding a paper on the other, looking at the camera with text on the paper saying "I am not real" that looks hand written and not in cursive. With the image quality looking like its taken in snapchat. Make the selfie extremely mundane and realistic with low contrast and saturation. The girl is wearing a sleeveless black flowery dress standing in front of the sink counter.
    3
  223. rubs hands intensively if I have time
    3
  224. i tried using GPT3 to correct my grammar once, didn't really work out
    3
  225. @Dempseyration thank you for your consideration tho!
    3
  226. There's a really in-depth walk through of Jukebox, might contain what you want to know https://youtu.be/1aO-uHXbzmQ You can also read the paper yourself, as it'll probs have the explanations
    3
  227. https://www.youtube.com/watch?v=xoWxv2yZXLQ
    3
  228. I just made an installation tutorial if you need it https://youtu.be/ICvNnrWKHmc
    3
  229.  @4.0.4  I get what u saying, but the sponsor starts at 1:45, the whole scenery changed, company logo displayed, it's like screaming a sponsor segment is coming already. Idk what else I can add tbh, but I am open to ideas. and it's normal for sponsors now to request to remove timestamps or progress bar to indicate it's a sponsor segment, on top of requesting the "sponsor" word only to be mentioned at the end of the segment. Unless YouTube updates the policy where ad segment must indicate "sponsor" at the start, then all contracts coming at me will have these requests, unfortunately.
    3
  230. check out DomoAI and start your vid2vid experiments: https://r.domoai.app/bycloud If you enjoy these types of compilations, let me know!
    3
  231. rip nearly
    3
  232. check out this guy's videos and blogs https://jalammar.github.io/
    2
  233. Most of my results for games are pretty terrible too. I think for a good result u need to have a clean background or a clear silhouette so the AI doesn’t fuck up the recognizing the face. I don’t know what the parameters for the colab demo is, but I installed and used it with default parameters. And all my realistic games characters did not work too. I think it’s rather picking the right input than choosing the right parameters
    2
  234. aite homie I'll try my best to make good contents :p
    2
  235.  @venadore  o my
    2
  236. I do have a video for how to run DaGAN locally, you can check it out https://www.youtube.com/watch?v=LmS22DoJNsA
    2
  237. 😭😭😭
    2
  238. i miss my FIRST comments (
    2
  239. I would still want to have my own creative control lol but the script is corrected by gpt4
    2
  240. I uploaded the wrong version but its too late I guess :/
    2
  241. She really does look like Mio hahah
    2
  242. Revolutionize your job search process with HubSpot's Free Resources: https://clickhubspot.com/wyt check out my newsletter: https://mail.bycloud.ai
    2
  243. unfortunately the project has not been updated since published
    2
  244. ok thanks for your suggestion
    2
  245. I just made an installation tutorial if you need it https://youtu.be/ICvNnrWKHmc
    2
  246.  @user-px4vs2yh7l  this video got out-dated in 2 days 😂
    2
  247. Check out HubSpot's Free ChatGPT Bundle! https://clickhubspot.com/jgv5 maybe I'll also cover ring attention and tree attention next time
    2
  248. thanks for the shoutout <3
    2
  249. no way, I didn't even realize. Thanks:)
    2
  250. no lol not even close
    2
  251. PULSE Face Depixelize Colab https://colab.research.google.com/github/tg-bomze/Face-Depixelizer/blob/master/Face_Depixelizer_Eng.ipynb
    2
  252. ​ @Naamiko  It is really confusing if you are not familiar with command prompt. If I decide to do so, I will post it on my video dump account. Here: https://www.youtube.com/channel/UCfg9ux4m8P0YDITTPptrmLg
    2
  253. one of their dev just confirmed the o in o1 stands for openai they aim to be less confusing, but now its even more confusing with their multi-modal gpt-4omni
    2
  254. lmao can't believe this video is already outdated in just 3 months
    2
  255. lmao I saw his premiere right after I posted too
    2
  256. its a 10/10 bottom feeder lol rip
    2
  257. I’ve seen some of your too! Really like your content and keep it up!
    2
  258. u just haven't found it:)
    2
  259. here https://colab.research.google.com/drive/1bwUnj-9NnJA2EMr7eWO4I45UuBtKudg_?usp=sharing
    2
  260. good idea, will do
    1
  261. I coulddd and there are so much more I wanted to talk about too, but I think it'll get too long for a "conceptual" video about RAG... thanks for letting me know tho
    1
  262. I put in the kid in God of War and it gave me a long haired girl lol
    1
  263. i used ffmpeg
    1
  264. brbrbrbrbrbrrrrrrr
    1
  265. no, i literally got goosebumps listening to his voice lol
    1
  266. tbh i made it so people needa pause them if they want to see it lol sorry if that's inconvenient but I doubt people would wanna see me read memes instead of showing cool AI news hahaha
    1
  267. im pretty sure they only said the image adapter is 100B for 405B params. As for the 90b and 11b, they didn't clarify how they have done it.
    1
  268. join my discord, its easier to talk there!
    1
  269. www.bilibili.com/video/av42596764
    1
  270. I put his twitter tag in the description but here's the link https://twitter.com/8co28
    1
  271. they just released some new quants https://x.com/AIatMeta/status/1849469912521093360
    1
  272. he has the cheat code
    1
  273. 👀
    1
  274. It uses this: https://github.com/mikedh/trimesh if that's what you are referring to
    1
  275. then there's Hent-AI and DeepCreamPy
    1
  276. I found the video by luck, but you can check out Neil Wong's stuff on reddit. https://www.reddit.com/user/neilwong2012 unfortunately rn reddit is dead so here's his twitter https://twitter.com/neilwong2012
    1
  277. I didn't train it myself, I downloaded a pretrained model
    1
  278. f
    1
  279.  @ShutterAuthority  there are a lot of apps you can find that does this, but some you would need to pay for them. This is from one of my previous sponsor if you're interested https://bit.ly/33qqySe
    1
  280.  @ShutterAuthority  this might be what you are looking for: https://youtu.be/is347MG71yY SimSwap requires 1 image to deepfake a video
    1
  281. This one is better: https://youtu.be/KRE4QZAXYu4 And I have a tutorial linked in the description
    1
  282. Feel free to contact one of my previous sponser 27 Stars, They can help develop your own AI software Check it out here: https://27stars.co.uk/bycloud Include "bycloud" in the initial email to receive a 10% discount for all your purchases.
    1
  283. no way, I hate my own voice lol I thought it would be irritating for people if I talk too long lmao
    1
  284. 0 words
    1
  285. check out the description
    1
  286. My own computer, GTX1080
    1
  287. it doesn't i think
    1
  288. already made my 28462846247335th edit in the script, maybe i’ll cover it later on 🥲
    1
  289. Well technically 1st place’s code is evolved from Deoldify. But I am not sure how exactly different, but it has definitely grew not only architecturally but also training wise.
    1
  290. I just made an installation tutorial if you need it https://youtu.be/ICvNnrWKHmc
    1
  291.  @F_E_U  that'll be sick actually, u mind dm me on discord?
    1
  292. im not exactly an expert but this is what I found https://en.wikipedia.org/wiki/Volume_ray_casting
    1
  293. Ya, tbh MusicLM's official demo is way too cherry picked. I tried giving the same prompts into Test Kitchen but the quality is not even close to their demo. Also one thing I found is that, Test Kitchen's MusicLM over extrapolate ideas while MusicGen stick to the given prompt more consistently. When you try just "ambient sound", MusicLM would give you a song, it feels like it was finetuned to generate songs, while MusicGen do actually give you ambient sound. In MusicGen's research paper, they stated that they used a research API for MusicLM to make their comparison. I guess they have a research only API which might be a more accurate representation of MusicLM? So their demo comparison seems pretty fair.
    1
  294. dude same
    1
  295. That was an amazing intro, love your content as usual Marques <3 There were a few misinformation in this video tho, sorry if I am being that person but I really just wanna clarify it! - Dalle 2(OpenAI) and Stable Diffusion (Stability AI) is not affiliated, Dalle 2 didn't use anything that SD provided or model related. The only thing that is similar is CLIP (the text embedding part) - Dalle 2 is trained on a custom YFCC100M flickr dataset (Creative Commons) which dalle 1 used, not LAION2B (well at least that's what they said on their research paper) Hopefully that clears out a bit of the confusion, but still an awesome AI video!
    1