This Open-Source Ai Video Generator Nailed It!

Okay Tim, so I want to start making 3d training videos with IClone instead of just 2d training videos with Vyond. What program will enable me to convert pictures of scenes or sets to .obj, or .fbx, etc. files for my Iclone videos without breaking the bank? (I'm an instructional designer).

This ai influencer tech has a new level now, google Fanvue and you'll see what I mean

I don't think the 1.3b can image to video if there is a link or place for such a workflow PLEASE correct me!

Initial reaction is that it sucks compared to Kling. But hey, it's free and is definitely good with that in mind. I get wild results that Kling NEVER gives me and the quality is blurry quite often, objects merge when cross each other.... although I may need to play with the Sampler settings more.

Testing it now. I'll let you know if it tops the king...which is Kling by a mile. One thin gI can say is that video generation is taking a long time on my 4090

I'd say that text generation example is rather odd; "WHERE Am I", with the 'm' being the only lowercase letter there, was that exact form asked for in the prompt? I'd expect "am" to be either lowercase or uppercase in its entirety. And the sign is oddly held, I'd think by looking at it that its just a flat sign, but one of the hands seems to be holding it by something behind it; does it have a handle of some kind, and if so, why is only one of the hands making use of it?

Still a very, very impressive open model, thx for your trademark substantial experimentation for a review!

It is no wonder Hollywood went on strike! Now anyone in the world has the potential to make a block buster movie!

Man!! There is so much happening so fast!! Its an exciting time! Thank you Tim for all of your help. Its very hard to keep up with all of this.

I attempted to run wan 14B on my home server with 3 24gb GPUs. I was able to initially inference the model but ran out of memory in the process. Has anyone heard of using something like GGML to quantize this bad boy to 8 bits?

Chinese AI is like...
Crouching Tiger🐯
Hidden Dragon🐲

666

Can you do a video on how to access Wan.21. Do you have to use ComfyUI?

Wanks!

Glad to see such high quality output from open source products! 👏🏻 👏🏻

I probably shouldn't even attempt this if I'm running a 3070, right?

I tried the Swarm UI for using WAN, man that was horrible. It had issues constantly, and the videos it did generate were poor quality. So, I'm starting over.

The cat! 🤯 I 'just' commented on one of your videos a few days (months) ago about kung fu in AI video! The dream is nigh! AI time is insane. When you started with 'It's new AI-video Model Day' I thought - it's noon already?! 🤣 Thanks as always, Tim. Sign the form please - "Keep it Wanx'.

Wan not free on Krea unfortunately

In india it shows coming soon.

For AI video I would appreciate if you start testing an anime fight scene from dragon ball z. I find they really struggle with that so it's a good test.

I clicked for bootleg Lara Croft.

haha is it just me or is wan's image to video running at 12-16 fps 🤣 it's completely unusable

Fantastic video! I personally don't mind the low frame rates because I can always fix that later, I'm just really happy to see open source video models competing with current SOTA models.

Well done, Lara! Arrows cost a lot to craft!!!

FAL $0.4 6min gen speed vs Replicate $0.2 20sec genspeed. So its good to shop around when using AI platforms. In runpod the 480p cost is about $0.04 for 5s.

Thanks for sharing! Amazing video

Hi Team, great video, how about voicing the guy in the blue business suit or the cyberpunk girl?

I think no one can beat google veo 2, google has a very crazy infrastructure, very big data, and a very big ecosystem too 🥶

Super annoying that luma thinks it should do background music in the action scenes. Can you promote not to do music?

Nope

Creepy doll phobia 😂😂😂😂

How to access this AI

I tried starlight I noticed it changed the faces of the characters a little bit… not sure if I like that every time … still good 👍😊

Wan can generate a 5-second 480P video on an RTX 4090 in about 4 minutes (without optimization techniques like quantization).

Just wait till people start refining the model and doing Lora’s within the next month

Cottageturtlecore 🤣🤣🤣!! That diving somersaulting cat is AWESOME!! Oh and every time you mention Tomb Raider, I want to shout at you...."It's LARA not LAURA!!!!!" 😂🤣😂🤣🤣😂 x

Well the free buffet didn't last long... 9h later, it eats credits on Krea. But impressive for sure. Anyone else has the problem on Krea that when credits reset, they are set at 75% computer? I really think they have an issue with how they count and this bug makes me feel like every month I can render less and less... their support isn't very helpful either

Krea only let me do one video for free then wanted me to upgrade to do any more.

Woa that "Cat fight" is hillarious. I gess if ones can find the desired footage in photostock it'll remain the most affordable. Then AI gnerated could help produce the exact clip the creator has in mind. Its still pretty far from stealing jobs of actors that actualy hold the marketing product in theyr hand...If occur, these software would be so expensive that I suspect you'd still make busyness with live actors marketing agencys...🤨...

We really are getting jaded about how amazing all of this actually is. Humans sure are something: we can learn to take anything for granted...

I think the creepy dolls would benefit a lot from some AI ghost voices, or high pitched screaming 😆

Wanx isn't free on Krea anymore, at least it's charging me compute units for the past 3 hours.

Thanks for your updated info every day. However, did you notice that image-to-video technology has noticeable updates regarding hands, physics, character consistency, facial expressions, and camera movement? There is still a problem with eye contact and looking direction or orientation.

Very great 👍 ❤