HellesGrind
@hellesgrind
Followers
74
Following
261
Media
56
Statuses
153
Aleksei, love gen ai stuff, founding engineer @ https://t.co/AUFae8T9pA backed by Sequia and First Round Capital
London, UK
Joined November 2022
A new solid AI news feed has just launched, covering what’s currently trending - from github, huggingface, reddit and @replicate
https://t.co/juJv3RoDHh I checked it out and can confirm - if you’re looking for truly fresh news, it’s there and it quickly makes its way to the top.
0
1
4
Maybe I’m the last one to discover this, but this Claude Code feature just sped up my dev workflow by ~30%. alias claude="claude --dangerously-skip-permissions" No confirmations for any operations — true YOLO mode. Use at your own risk, ofc 😄
1
0
0
I'm the type of person constantly looking for different mac OS tools to enhance my experience. Here's a list of what I use in my day-to-day life: cursor + claude code cli obsidian - my task log/thoughts/lists google calendar + tasks - for my scheduled tasks and meetings cleanshot
1
0
0
tried this prompt for ehnancing realism of AI photos, shared by @gokayfem but on Seedream-v4.5 Adds more realistic texture/imperfections to skin and clothes details (right one with a prompt) imo the best model for AI character portraits is Seedream-v4.5. Add this to your prompt:
0
0
1
AI Workflow for dancing videos using Kling Motion Control: 1. get selfie of yourself and generate a full height 9:16 photo using Nano Banana Pro (I made mine on @fal) 2. Find some cool video on YouTube 3. Animate photo using reference video and photo (@fal or @higgsfield_ai)
1
0
4
btw i had no proper full-height photo for this test, so i just used nano-banana pro to help me out
0
0
0
I tried Kling Motion control and it's so damn cool! Face consistency is a bit off, but camera and motion dynamics are sooo good. (actually never thought I'd enjoy making dance videos w/ myself so much) Have a great Christmas! Some more examples in🧵
2
0
0
Alibaba released a new model which might be a good addition to Wan Animate - ViSA. Really like the 3D aware feature - really making body and movements much more realistic. Not available via API yet, but you can find a paper here: https://t.co/BQMEJIHB64
0
0
1
Definitely will check it out!
ViSA: 3D-Aware Video Shading for Real-Time Upper-Body Avatar Creation Contributions: i) We propose a novel framework that integrates one-shot 3D avatar reconstruction with a real-time autoregressive video shader, unifying geometric stability with generative rendering power. ii)
0
0
1
Steady Dancer - new model for UGC dynamic videos. While One-to-All is not available on API (@wavespeed_ai @fal - waiting for it!) people on X say that this model is good. Looks a bit AIish to me, wdyt lol? You can try it on wavespeed (link in thread)
1
0
1
New release that could become the go-to model for UGC-style AI video: One-to-All Animation. Refreshing to see releases in UGC video space. Cinematic video gen gets weekly drops (Kling 2.6, etc), but motion transfer models rarely get updates. This is the video-driven animation
1
0
3
It’s insane how fast models start to feel so damn OLD. Just couple of days using Opus 4.5 and Sonnet feels magnitudes dumber
0
0
0
Connecting Claude Code to my Obsidian vault genuinely made note-taking better for me. Not in a productivity-guru way. Just: I actually write stuff down now. The reason is dumb but it works — every note I write gives AI more context about me, so it becomes more useful. Now I want
0
0
3
Sooo good
Some people make noise on the internet for the sake of noise. But @soren_iverson knows how to make it mean something. This is how the Weird Al of tech turned his absurdity into a world-class design agency. Full episode on YouTube.
0
0
0