DΞV

@junwatu • 5,083 subscribers

Founder of Junwatu. Sharing AI engineering and creative experiments.

Shorts

This is what you get running H3 on a single H100. Stack: - ComfyUI + official H3 T2V workflow - H3 ComfyUI pruned BF16 - GPU H100, 80GB - 768p, 10-second video cost $1 - Average generation time: ~14 minutes This kind of deployment makes sense if you want to build and run your own system or custom services. If you don’t want to bother with the infrastructure, just use an API. Generated this video for testing.

This is what you get running H3 on a single H100. Stack: - ComfyUI + official H3 T2V workflow - H3 ComfyUI pruned BF16 - GPU H100, 80GB - 768p, 10-second video cost $1 - Average generation time: ~14 minutes This kind of deployment makes sense if you want to build and run your own system or custom services. If you don’t want to bother with the infrastructure, just use an API. Generated this video for testing.

51,014 просмотров

When FLUX 3 goes public, I think we could see more AI filmmakers. Open weights model lower the barrier to creation, giving more people the chance to experiment with storytelling. With video generation still being expensive today, that's a big deal!

When FLUX 3 goes public, I think we could see more AI filmmakers. Open weights model lower the barrier to creation, giving more people the chance to experiment with storytelling. With video generation still being expensive today, that's a big deal!

11,295 просмотров

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

This is only possible by using an existing video. Below is the source video, Russian Roulette by Kotte Animation. The bigger picture here is how easy it has become with AI workflows to take an existing scene and completely reskin it while keeping the original motion and choreography. Yes, bypass all the complex prompting.

This is only possible by using an existing video. Below is the source video, Russian Roulette by Kotte Animation. The bigger picture here is how easy it has become with AI workflows to take an existing scene and completely reskin it while keeping the original motion and choreography. Yes, bypass all the complex prompting.

577,447 просмотров • 1 месяц назад

If AI slop is like this, please keep making it! Never thought that Wan 3.0 would be this good at slice-of-life scenes. FlovaAI x Wan 3.0 Limited-time offer: Wan 3.0 480P starts at $0.013/second.

If AI slop is like this, please keep making it! Never thought that Wan 3.0 would be this good at slice-of-life scenes. FlovaAI x Wan 3.0 Limited-time offer: Wan 3.0 480P starts at $0.013/second.

594,289 просмотров • 1 месяц назад

MATIIN IKLAN ANDROID Iseng iseng konekin hp Android ke laptop dan minta agy mattiin ads. Ternyata semua apps dan services yg handle iklan di disable sama agy. MAMPUS!! 🤣🤣🤣🤣🤣🤣🤣🤣 Saya greget bgt soale dri dulu gk ada cara efektif buat matiin iklan-iklan yg tiba-tiba muncul di hp Android. Cara ini juga bisa buat hapus atau disable bloatwares. Nih, videonya ⬇️

MATIIN IKLAN ANDROID Iseng iseng konekin hp Android ke laptop dan minta agy mattiin ads. Ternyata semua apps dan services yg handle iklan di disable sama agy. MAMPUS!! 🤣🤣🤣🤣🤣🤣🤣🤣 Saya greget bgt soale dri dulu gk ada cara efektif buat matiin iklan-iklan yg tiba-tiba muncul di hp Android. Cara ini juga bisa buat hapus atau disable bloatwares. Nih, videonya ⬇️

97,268 просмотров • 4 месяцев назад

H3 generates 10s in under 3 minutes 🤯 After adding LoRA, the generation time dropped from 14 minutes to just under 3 minutes. God damn! Stack: - H100, 80GB - ComfyUI latest - minimax_h3_fl2va_pruned_bf16 (40GB) - Kijai/MiniMax-H3_comfy on Hugging Face - PyTorch 2.13 + CUDA 13.0 The video quality is great, but the audio really needs some processing. All video resolution is 768p.

H3 generates 10s in under 3 minutes 🤯 After adding LoRA, the generation time dropped from 14 minutes to just under 3 minutes. God damn! Stack: - H100, 80GB - ComfyUI latest - minimax_h3_fl2va_pruned_bf16 (40GB) - Kijai/MiniMax-H3_comfy on Hugging Face - PyTorch 2.13 + CUDA 13.0 The video quality is great, but the audio really needs some processing. All video resolution is 768p.

46,760 просмотров • 2 месяцев назад

Asked FLUX 3 for a cinematic watermelon slicing scene. It went harder than I expected. 🍉⚔️

Asked FLUX 3 for a cinematic watermelon slicing scene. It went harder than I expected. 🍉⚔️

17,449 просмотров • 2 месяцев назад

FLUX 3 is really good at creating old footage.

FLUX 3 is really good at creating old footage.

12,056 просмотров • 3 месяцев назад

Instal OpenClaw 🦞 Ini seri video ya, ASMR. Target beginner. Silahkan di bookmark kalau membutuhkan. 1️⃣ Setup OpenClaw Setup cukup mudah tinggal pake satu command dan akan meng-instal semua software yg dibutuhkan. OpenClaw juga menyediakan yg namanya antar muka chat melalui terminal (tengah akhir video). Jadi gak perlu buka web. Tergantung setingan model AI, kalau bagus banget bakalan mengerti banyak bahasa. Di sini saya pake MiniMax M2.1

Instal OpenClaw 🦞 Ini seri video ya, ASMR. Target beginner. Silahkan di bookmark kalau membutuhkan. 1️⃣ Setup OpenClaw Setup cukup mudah tinggal pake satu command dan akan meng-instal semua software yg dibutuhkan. OpenClaw juga menyediakan yg namanya antar muka chat melalui terminal (tengah akhir video). Jadi gak perlu buka web. Tergantung setingan model AI, kalau bagus banget bakalan mengerti banyak bahasa. Di sini saya pake MiniMax M2.1

16,523 просмотров • 7 месяцев назад

Больше нет контента для загрузки