Havard Axel (@HavardAxel) profile picture

AI builder https://teapawt.net/t.co/U0g6BwbHiZ https://teapawt.net/t.co/C11R818z6J

Joined February 2020
Photo by Havard Axel (@HavardAxel)Photo by Havard Axel (@HavardAxel)Photo by Havard Axel (@HavardAxel)
Fake AI reviews to attack competitors and make themselves look good, this is what the internet has come down to with brands like @magnific . The game is super biased, ChatGPT uses these trustpilot reviews to give advice on which tools to use.
1
1
69
Let's see if this decision turns out as good as for @jackfriks 's @postbridge_
Uwear is now an official plugin in @ChatGPT, just in time for the new desktop experience!
2
15
Havard Axel retweeted
Introducing Uwear MCP. The first MCP dedicated to fashion visual production. Connect it to Claude or ChatGPT and run photoshoots like a multi-million dollar brand: ✅ Access frontier AI image and video model like Nanobana, Seedream, Kling and more. ✅ Give your agent fashion-native building blocks to work with: garments, outfits, models, and art direction to make every look unique and consistent. ✅ Generate 1 to 100,000 visuals from the same chat. ✅ Every asset and result lands in your Uwear workspace, or in your local folders from a desktop app Ask your agent "install Uwear from uwear.ai/mcp"
1
1
1
2
246
Hey ! With my team we are working on github.com/Illospace/illospa… (open-source multi-player openclaw-like). We are wondering if we should build an app for it or keep it browser based, any opinion ?
1
2
23
Havard Axel retweeted
🚨 GPT Image 2 is live on Uwear day 0! Can it dethrone the king, Nanobana Pro, for fashion photoshoots? We put them side by side on three tricky garments that aren’t easy to reproduce. Which one would you pick?
1
1
2
131
Havard Axel retweeted
AI product photography is powerful. Setting it up AT SCALE is not easy. Introducing Agent Mode and Mouza: the last AI fashion photographer you'll hire 🧵
1
1
1
1
46
We should be able to talk over Claude Code. I use Claude Code daily, and I can see when it starts going in the wrong direction. However, there is no way to tell it to stop and redirect its thoughts. Somehow I have to wait 15 minutes of nonsense or risk breaking the context.
1
1
33
Current LLM platforms don't feel like great conversation interfaces
11
Havard Axel retweeted
I always felt CFG was a patch to fix a training problem we didn't yet understand. Training with only normal distributed noise teaches the model that each step will have a perfectly normalized error from the previous step, which is not the case. Therefore, it is incapable of correcting the errors it created from the previous steps during generation, which leads to distorted generations as these uncorrected errors compound with each step. We currently correct this by applying 2 pass CFG to amplify the model's correction predictions from a base, which helps correct these errors at each step, but this leads to the model over correcting leading to over corrected images that look oversaturated. That classic AI look. I tested fine-tuning Z-Image while providing a balanced random augmentation of the noise and it appears to have taught the model to overcome these errors which led to the model no longer needing CFG and also producing better quality images in the process. These samples are from training a LoRA on Z-Image with a batch size of 2 for 3,000 steps. I am going to do a significantly longer fine-tune using the same process this weekend.
23
27
4
403
26,139
Havard Axel retweeted
seedream 4.5 is one of the first model capable of consistently generate completely new photos 4 ref of clothing with a hoddie/vest on top one human model My face is weird because I am making a weird face in the input photo 😂😂
2
4
413
Havard Axel retweeted
🍌 Nanobana Pro is now on your favorite AI Fashion Studio 👗 It is the best image model currently available and we just tuned it to give out the best fashion photos! ✨ For 12 hours: follow, like, retweet & comment = FREE CREDITS to try it out ✨
1
2
2
57
Havard Axel retweeted
🎉 New App for shoppers is live on App store! Create your avatar, get a color analysis, try clothes and share with friends. This app builds on 1 year+ of learning operating @uwearai, an AI photo platform which is used by hundreds of brands of all size today.
1
2
1
2
754
Havard Axel retweeted
Drape1 is back and FREE to use on uwear 🚀 It's not as accurate as Drape2, but still works great for basic clothing and is 60x time faster than chatgpt 😉
2
1
70
Havard Axel retweeted
This is probably the best AI workflow for editing. But for a flat-lay to on-model? Not as good nor fast as Drape2 on @uwearai ! One flat + one prompt + 5 sec = 712x1024 output with SOTA accuracy
Replying to @OpenAIDevs
Photoroom is using gpt-image-1 to help online sellers instantly create studio-quality visuals, lifestyle scenes, and on-model shots from a single product photo.
2
5
73
Havard Axel retweeted
"Little boy wearing a hockey jersey, in a parc, on a sunny day" Shoot hundreds of clothing in minutes only on #uwearai #AIphoto #showcase #photoshoot #fashionai #ecommerce #Drape2
1
2
21
Havard Axel retweeted
🚀 New Feature Alert ✨ Enhance - lets you add realism and details. Great for situations where clothing looks perfect but the background or face needs adjustment. ↕️ Upscale - doubles the image resolution, from 768x1024 to 1536x2048, which makes the images clearer and sharper.
1
2
2
51
Havard Axel retweeted
🎉We are thrilled to announce the public launch of Uwear.ai and our core AI model Drape v1.🎉 With Drape, you can generate on-model images with state of the art accuracy, using only a flat-lay photo of your item, in less than 30 seconds. Links in the comments 🔗
1
1
1
3
264