OMG! @Kling_ai 4.0 is OUT TODAY !!!!
I made this musicvideo with it! Here is everything you need to know …
This post is sponsored by Kling AI but they allowed me to create whatever I wanted!
I was one of the first few artists who got early access and used it to make this music video.
What stood out to me most wasn’t just the image quality, but its ability for lipsyncing!
For the lip sync, I used my preferred method: rendering the actual music track with a black frame and combining it with the stills I created.
Same with prompt adherence.
Shaky handheld camera actually felt shaky.
The anamorphic lens look came through.
And with Omni Reference, you can now work with up to 15 references across images, video, voice and elements.
On launch you also get:
• up to 10 keyframes
• 30s generations
• 1080p
• multi-shot support
• improved lip sync
• improved text generation
• much more flexible reference control
Kling 4.0 feels less like “generate me a video” and more like “here are all the ingredients, now direct it.”
oh, and its cheaper than seedance! And 4.0 Flash is A LOT faster too!
Have you tried it? Let me know your thoughts!
Thank you so much for helping me with this one Ben Chimoy!
🇸🇬 🇬🇪 @tether, @pave_bank and @gftn_official launch the Georgia-Singapore Fintech Scholars Programme.
Top students from Tbilisi State University spend a semester at Singapore Management University, fully funded.
The first higher education exchange between Georgia and Singapore.
@pushinpixls Yeah, There's this dumb law (PSD2) that says the business of providing APIs is regualted and you're supposed to use an authorized TPP for all your bank access (like truelayer). We (monzo, im not there anymore) took a bit of regulatory risk keeping the developers API live FYI.
Don't sleep on @modal . The feeling of spinning up 8 H100s for 10 minutes to batch process a bunch of stuff and having it come in under the monthly free tier is great.
It used to be a faff to make sure your workload could be massaged into a python function, but claude knows how.
I tested this with a text based name classification task and it performed badly. Question was something like "Given this user name and country of residence, does this look like a valid individual name. Look out for things like OCR errors, indications the name may be fake or a company name."
@stevegraham@ChatGPT I used Computer Use for the first time the other day. Needed screenshots for some audit evidence.
Walked away and 20m later it was done. Just perfect. Instantly upgraded.