Generating content with Local AI - 4070Ti Super 16gb/32 GB RAM. ex (Freelance programmer, 2d/3d content creator, Tech author & Game developer)
Kochi, Helsinki
Joined August 2008
- Tweets23.6K
- Following366
- Followers994
- Likes2.9K
Pinned Tweet
Vaaripanjara - The Cage Of Water S0E0
Starting #AI #microdrama series with 1min episodes
Generated locally on a #4070Ti with @krea_ai & @MiniMax_AI #MiniMaxH3
Story loosely fuses Indian mythology with regional urban legends
Subscribe & follow the story
youtube.com/shorts/iG7M4_RJS…
Vaaripanjara #Microdrama Episode 8
Nisha gets interrogated 🧐
Next episode will reveal her identity 😎
Made using @ComfyUI with #MiniMaxH3 on my 4070Ti ☺️
youtube.com/shorts/mlEyb2CdS…
1
1
21
Vaaripanjara #Microdrama Episode 7 😇
This one took a lot of effort. 50+ takes. 🥴
Happy about the result 🤓
Made using @ComfyUI with #MinimaxH3 on my 4070Ti 😎
youtube.com/shorts/JfvYCw4Un…
36
Vaaripanjara #AI #microdrama Episode 6 is out 🔥 🌶️
This one ties everything together to make sense of what happened so far. 🤓 Mostly
Made using @ComfyUI using #Krea2 & @MiniMax_AI #MinimaxH3
Next one is a tough one though 😇
youtube.com/shorts/hlT4q1m-H…
59
Vaaripanjara #AI #microdrama Episode 5 is out ☺️
The Villain Emerges 😎
Need to investigate more on refmods to integrate to the workflow.
Made using @ComfyUI with Krea2 & MinimaxH3 on my 4070Ti 🤗
youtube.com/shorts/P7YcDiWRh…
54
Vaaripanjara #AI #Microdrama series Episode 4 is out ☺️
Did notice those lip movement overlap issues, were due to upscale pass reinforcing the prompt. But need to keep the cadence, so onwards 🤗
In this one a new team gets formed.
youtube.com/shorts/yQ7RqATex…
1
53
Vaaripanjara #AI #Microdrama Episode 3 is out 😎
Turn around has become much faster taking under 6hrs for an episode on my 4070Ti 🤓
Made with love using @ComfyUI
with MinimaxH3 & Krea2
youtube.com/shorts/iWJK0eyUg…
1
1
150
Vaaripanjara Episode 2 is out 🔥
Things getting spicy 🥵
Switched to Obvpm Minimax workflow from pure motion context.
Made with @ComfyUI locally with a 4070Ti 😎
I am proud of how this turned out in limited time. Thanks again to Krea2 & Minimaxh3 🤗
youtube.com/shorts/9d_Wh_QCr…
1
100
Great stuff @PrunaAI Thanks for this specialised Minimax model 😊
Wish I could get my hands on the system prompt for the prompt upsampler for the editing workflow 🤗
Another image gen model added to the collection - @PrunaAI p-H3-Image
Image shows results with same prompt, seed & sampler across all of them. Nice to have variety.
Testing to see how we can use references to generate scenarios. Minimax model just keeps on giving. GG
29
Vaaripanjara - Episode 1 of my #ai #microdrama series is out ☺️
Created with @ComfyUI locally on my 4070Ti using @krea_ai & @MiniMax_AI
Loosely based on Indian mythology & folktales. Do subscribe 😄
youtube.com/shorts/Ww6XUAprn…
1
174
Happy Kerala Piravi wishes 🤓
Legend has it that the state of Kerala was created from the sea when sage Parashurama threw his legendary axe across Gokarna to Kanyakumari
Made with Krea2 & MinimaxH3 overnight on my 4070Ti 😊
youtube.com/shorts/h9-zsHbbT…
1
1
135
Found this MinimaxH3 based character sheet generator, which works properly.
github.com/shisa84/ComfyUI-H…
Tutorial
youtube.com/watch?v=HevfhJnV…
Made some adjustments to get it the way I like my sheets.
1
86
JuwalBose retweeted
Meet Pika H3 Flex, debuting today as the #1 open weight model on DesignArena’s t2v board.
It’s built for professional iteration, with exportable audio stems, resolutions up to 4k, and faster-than-realtime generation. It can produce a 15-second clip with soundtrack in as little as 7 seconds.
So, yeah. It’s kind of a flex 💅
JuwalBose retweeted
Agaram 2.5B is out. It's a Tamil-first language model we trained from scratch at @MuraiLabsHQ , our small non-profit lab. As far as we know, it's the first Tamil model trained from scratch that can call tools.
It chats in Tamil, Tanglish and English, translates, and can call tools. We're releasing the base model, chat model and GGUF files under Apache-2.0. You can download it and run it offline on your own laptop.
The first chat model had a four-day public preview. Seeing where it fell short with real users helped us decide what to work on. We rebuilt the fine-tuning around what we saw, using 222,662 records.
Thank you to everyone who spent time with that early version.
The video walks through the model. You can try it here:
hf.co/spaces/mechramc/agaram…
நன்றி 🙏
JuwalBose retweeted
I did a quick test on Qwen3.8-Flash-Next (125B, Gyro-S, @support_huihui abliterated) local on a 16 GB 4090 laptop + 32 GB RAM via @coldniko's Strata.
75–110 tok/s gen, 1,949 tok/s prefill, 262K context.
Frontier-hard: 10/11 math/code, 7/8 vision, 2/3 Blender 3D (+1 partial). Missed niche facts.
JuwalBose retweeted
🚀 Open-H3-Regenerate-2K: 3-Step Video Regeneration, Native Output up to 4K
MiniMax open-sourced H3 Base for 768P video generation, but its official high-resolution regeneration stage, H3-Regenerate-2K, remains API-only. We provide an open alternative, bringing high-resolution H3 regeneration to your own GPUs.
⚙️ Just 3 diffusion steps, regenerating video directly at the target resolution
🎬 Native output up to 4K, supporting 768P, 1080P, 2K, and 4K for 5–15-second clips in one pass
🔌 Plug-and-play Refiner LoRA trained on H3: load the weights into H3 Base and start refining, with no retraining required
⚡ 5.0× speedup over the official API: for a 5-second 2K clip, the official API takes 112.7s; ours takes just 22.45s on a single B200, powered by SoL-Engine and Fast H3 VAE
📝 Blog: nvlabs.github.io/Sana/Open-H…
💻 Code: github.com/NVlabs/Sana/tree/…
