Jeff Tang (@JeffereyTang) profile picture

Founder of LibAI, the creator of https://nitter.gq/t.co/zWSrgetVgU, https://nitter.gq/t.co/Y0CvPgA9TC, https://nitter.gq/t.co/K9CjLM1Z7q and https://nitter.gq/t.co/qGE6EPdN6N

Joined January 2017
Photo by Jeff Tang (@JeffereyTang)Photo by Jeff Tang (@JeffereyTang)Photo by Jeff Tang (@JeffereyTang)Photo by Jeff Tang (@JeffereyTang)Photo by Jeff Tang (@JeffereyTang)Photo by Jeff Tang (@JeffereyTang)Photo by Jeff Tang (@JeffereyTang)
This week every lab shipped the "almost as good, way cheaper" model. Gemini 4 Neon, GPT-6.2 Sol. Image tools have the same split. A 100ms background-removal model handles most product photos fine. The hard ones, a logo the same color as the backdrop, a frame that fills the whole shot, need a model that actually understands what the product is. I don't think the answer is picking one. It's routing: fast model by default, big model only when the small one isn't sure.
35
A question for people who actually cut out product photos every day: If a much stronger matting model showed up — DiT-based, around 10B params, maybe 10 seconds per image, but with real semantic understanding — would you use it? I mean the cases small models still miss: a logo whose color matches the background, or a large picture frame with only a sliver of backdrop around it. The model would know what to keep, not just what looks like an edge. Or would you still take the 100ms model that's maybe 10–50x cheaper, and live with the hard cases? Curious where people actually land when quality, speed, and cost are all on the table.
1
13
SSENSE is getting called out for AI product photos where zippers move, embroidery flips, and details change from shot to shot. That's the line for AI in ecommerce: restyle the scene all you want, but never redraw the product. Keep the real pixels of the item. Generate everything around it.
1
41
Seedream 5.5's real upgrade isn't quality. It's native RGBA. When a model can emit a clean alpha, generate → cutout → listing starts to become one step. We built that last mile at cutout.pro. This is the first time those two worlds feel like they're meeting.
1
1
29
I am the founder of LibAI, the creator of several AI products with great traffic. If you are interested in building and growing AI products together and exchanging tech and GTM ideas, let's connect. reply to @JeffereyTang
13
10
211
We just got access to Seedance 2.5 in native 4K. This isn't just more pixels. It's cinematic-grade output — the kind of image quality film creators have been waiting for in a generative pipeline. For filmmakers, this is a real step: resolution and look that can actually sit in a cut. If anybody needs access, please reach out.
2
59
Something clicked for me this week. Higgsfield shipped a tool that vectorizes a finished video into layers, so you can change one object without regenerating the whole shot. OpenAI’s new image models can export SVG. We’ve been so obsessed with generation that we forgot the boring, valuable part: making the result editable.
1
1
26
remove.bg’s standalone site is winding down (consumer BG removal moves into Canva; Dec 1, 2026). If you just need a clean cutout on the web: cutout.pro still keeps background removal free for regular users — no “migrate or lose the tool” drama. We also shipped a big cutout model upgrade yesterday. Faster, cleaner edges, more robust on messy real-world photos (hair, products, the usual pain). Try it → cutout.pro
1
2
94
remove.bg’s standalone site shuts down Dec 1 — background removal moves into Canva. If you just need a fast free cutout (transparent PNG, no watermark), try ours: cutout.pro Built for product shots & creatives who don’t want another subscription maze.
56
Everyone's shipping 15-second AI videos this week. Fair. Meanwhile most ecom sellers still lose hours on messy product shots and SVG exports. The boring layer — cutout, cleanup, vector — is where daily throughput actually compounds. We built that for free at cutout.pro. Curious what still slows your listing workflow?
1
2
32
We made the content. You post it. You get paid. Cutout Creator Program: 1️⃣ Pick a ready-made post 2️⃣ Share it on TikTok, YouTube, IG, FB or X 3️⃣ Submit the link — earn up to $8 No filming. No editing. Just 50 followers. Closes Nov 4–5 👇 cutout.pro/creator-program
4
129
We shipped image vectorization on Cutout — built in-house, and it beats Vector Magic Quick product note from us. We just put image vectorization live on Cutout. It’s our own model — not a wrapper around someone else’s API. We built it because POD, design, and e-commerce teams kept asking for the same thing: turn a raster into clean SVG without babysitting the output. In our tests it’s faster than Vector Magic, and on messy real-world images (logos, product shots, line art) the edges and small details hold up better. Try it here: cutout.tech/square/experienc… Drop a before/after if you try it — especially anything Vector Magic usually chokes on. We’re iterating from real files, not demo screenshots.
1
46
AI is becoming a productive tool not only for design, but also for marketing. #realestateagent #promeai
2
96
The biggest lesson we’ve learned building Ideal House 🏠: The pretty AI room image isn't the product. It's just the hook. The real product is the invisible engine behind it: 📦 Syncing real merchant SKUs 🔍 Previewing exact rugs, tiles, or furniture 📏 Respecting true size, material & variants 📊 Tracking what shoppers actually care about 🛒 Integrating seamlessly into PDP workflows We aren’t building just another image generator. We’re building a catalog-to-visual-commerce workflow. Pretty images win attention. Reliable workflows win adoption. 🚀
1
55
The popular post. How many of them did you follow?
49
The absolute biggest "dark horse" in the AI video space right now is HappyHorse 🐎! It just parachuted straight to the top of the video arena leaderboards, completely disrupting the space with its native audio-video capabilities. 🤯 A quick look under the hood: 🔹 15B parameter single-stream Transformer 🔹 Lightning-fast inference: 1080p in just 38s on a single H100 🔹 Native lip-sync support for 7 languages 🔥 The ultimate flex: Fully open-source & commercially permissive! We've been deep in the trenches building multi-modal video generation pipelines, and an open-source model of this caliber sends a massive shockwave. The traditional bottlenecks for producing AI short dramas—high rendering costs and strict audio-visual alignment—are practically being leveled overnight. What’s your take on this open-source surge? Is HappyHorse the "iPhone moment" for AI filmmaking and short dramas? How do the closed-source giants defend their moats from here? Drop your benchmark thoughts and hands-on experiences below! 👇💬 #AIVideo #HappyHorse #OpenSource #GenerativeAI #AIFilmmaking #TechTwitter
1
1
157
The rumored Manus acquisition — and reports that it may be blocked — raises some interesting questions. If a country chooses to restrict such a deal, is it primarily about protecting strategic technologies and data? If the deal is allowed, is it about enabling founders to access global capital and scale innovation faster? As AI companies become more foundational, where should the line be drawn between: national interest, open markets, and technological progress? And more broadly — are we entering a phase where every significant AI company is no longer just a startup, but something closer to infrastructure? Curious how others are thinking about this.
66
DeepSeek V4 might be the most polarizing release in AI this year. On one side: “Open-source SOTA.” “Blazing fast reasoning.” “Insane cost efficiency.” “Engineering masterpiece.” On the other: “Underwhelming in real use.” “Hallucinations still a problem.” “Hype > reality.” So what’s actually going on? We might be missing the bigger picture: DeepSeek V4 isn’t trying to win benchmarks — it’s trying to win the deployment layer. Speed + cost + decent reasoning perfect reasoning that no one can afford to run That’s a different axis of competition. But here’s the uncomfortable question: If a model is fast, cheap, and “good enough” — does it matter if it’s truly SOTA? Or are we entering the era where engineering beats intelligence? Curious where people land on this 👇
1
1
105
will be testing its coding ability
47