So far, the best Pika tutorial I’ve found on TH-cam. Subbed! You tackle the exact issues I’ve been having. Love your simple style and no frills, stupid jokes or fancy editing. Just the good stuff.
Love the direct instructional. Thank you! I have to say, Pika and some of the others I've tried are fun. But they are a time suck. I hope the soon to be Pika 1.0 has features that bypass the need to extract the last frame. A simple "Continue From Here" button would be a HUGE jump forward. As of now, I can spend 4 days barely getting anything usable. (I'm trying to my own videos instead of relying on cheesy stock footage). God bless all the others who are making videos over 2 minutes. Please keep us updated.
I haven't used Runway that much, from what I've seen it does get more dynamic (cinematic) camera motions and head movements for people. Pika allows for a bit more creativity in your prompts and is frequently updating with new features. Both are still in the early stages of development and have a lot of room to improve.
As far as I can tell the results are the same. The only difference is if you use /create you have to enter a prompt, whereas for /animate you only need a reference image.
Why not use ffmpeg to extract the final frame? I really hope PIka Labs creates an API soon. Then we could automate making longer videos and also batch attempts using a basket of different gs/fps/motion parameters so we don't have do that process by hand. Thanks for the prompt tricks, I have been having trouble getting the right motion out of Pika at times. I beat my head against the wall trying to get it to generate a car moving left to right in front of the camera, but never succeeded. I'm eager to try the "motion tracking" phrase you mentioned along with your other tips.
Any software that can get the final frame would work fine. ffmpeg is certainly an option. Pika labs does seem to be updating fairly frequently so I do think they will allow you to generate longer motions soon. I hope you have better luck with the motion, it seems easy to get more movement with the newer updates, but it is hard to control.
I have a problem that the lightning doesn't align, it's a slight change, these are my parameters visible eye blink, visible chest breathing, visible hair blowing, - camera wiggle, -neg morphing, erratic fluctuation in motion, noisy, bad quality, distorted, poorly drawn, blurry, grainy, low resolution, oversaturated, lack of detail, inconsistent lighting What should I do?
What do you mean by the lighting doesn't align, are you using a reference image? I see in your prompt you have "-camera wiggle" , for the -camera parameter it can only do zoom {in, out}, pan {left, right, up, down} or in a combination of them, and rotate {clockwise, counterclockwise}.
Hello, I have two questions to ask: 1. Capturing the last frame of the video may result in a color difference between the second video and the first one, as the captured image resolution is already compressed twice. 2. When merging the first and second videos, there are always inconsistencies before and after, such as pauses or inconsistent content Do you have a good solution?
Hey, for your questions unfortunately Pika labs is still pretty limited when it comes to extending the length. This video shows a potential workaround, but you'll still run into issues as far as inconsistencies between the first and second videos. I'll post an update video if they extend the video generation length though.
i tried this method but the generated video is not consistent in color and contracts and even has different detailed for example the eyes and facial features. i had quite a few iterations but the result is not desirable. how should improve this?
It's hard to get this to work perfectly, Pika Labs is still limited in generating high quality long videos unfortunately. Perhaps changing the prompt to be a bit more descriptive in the details you want and using a high guidance scale parameter (like "-gs 20") could help, but there will still be some warping compared to exactly what you're original reference image was.
Thank you so much you solved my making pikalabs longer videos.. would you also tekl me which text to image would be good for creating mlre refrences images which should be also free .. there are two i knkw. First is Ideogram and bing dalle 3.. which one would you suggest.. i know you might be tell Leonardo
Ideogram and Dalle3 is a good start. I think Adobe Firefly is free atm which is worth looking into, I know they just updated their models. If you are more technically inclined you can always use Stable Diffusion which I think is the best for free users, because of all the open source development being done and no censorship, although it has a high learning curve.
@@taoprompts thnx for the deep reply.. by the way.. I find Ideogram always works well this is the only which is free adm no limitation. Daale has limitations.. and everything else is not free
Some very important points about there about motion tracking... Do you have some kind of document like a PDF that has also these terms I could use for my projects?
Underneath the waitlist there is a set of links that says "Twitter Discord About us Careers". Click on that discord link and you should have access to the old beta version.
As far as I know, it's possible to generate videos without watermarks in a special discord channel, but you need to reach out to pika labs social media team or discord people to get access.
In the "faq" channel on their discord, there's a section for how to get non-watermarked videos. You need to contact their social media team to get the info on how.
Here's the UPDATED Guide for PIKA 1.0! th-cam.com/video/F4vZTKOCeds/w-d-xo.html
So far, the best Pika tutorial I’ve found on TH-cam. Subbed! You tackle the exact issues I’ve been having. Love your simple style and no frills, stupid jokes or fancy editing. Just the good stuff.
Thank you! I'll make follow up videos for Pika when they update.
@@taoprompts Sweet!
wow this is some big brain work around to pika's limits, thanks!
Great tutorial, Tao! Very useful.
Great information, please keep doing more of them
Thank you!, I'm working on videos for the new Pika 1.0 Ai now.
waiting for it, thanks alot@@taoprompts
Fantastic video, was short and to the point. 👍 Subscribed
Thanks for the support!
Bro looove how clear you are! Will definitely be using your channel! Thank you!
Thanks man, Im doing my best!
Love the direct instructional. Thank you! I have to say, Pika and some of the others I've tried are fun. But they are a time suck. I hope the soon to be Pika 1.0 has features that bypass the need to extract the last frame. A simple "Continue From Here" button would be a HUGE jump forward. As of now, I can spend 4 days barely getting anything usable. (I'm trying to my own videos instead of relying on cheesy stock footage). God bless all the others who are making videos over 2 minutes. Please keep us updated.
We should have access to Pika 1.0 really soon. They do need an extend video feature like Runway ML, I'll keep you updated!
Very simple to learn! You earned my subscription. Thank you! I'm new to Pika Labs.
Fantastic video. thanks for all your pika tips
Thank you, and you're welcome!
Super informative! 💥Thanks! What is your experience with Runway, how does it compare to pika in your view?
I haven't used Runway that much, from what I've seen it does get more dynamic (cinematic) camera motions and head movements for people. Pika allows for a bit more creativity in your prompts and is frequently updating with new features. Both are still in the early stages of development and have a lot of room to improve.
great tutorial. what is the difference between /create +reference image and /animate+reference image?
As far as I can tell the results are the same. The only difference is if you use /create you have to enter a prompt, whereas for /animate you only need a reference image.
Why not use ffmpeg to extract the final frame? I really hope PIka Labs creates an API soon. Then we could automate making longer videos and also batch attempts using a basket of different gs/fps/motion parameters so we don't have do that process by hand. Thanks for the prompt tricks, I have been having trouble getting the right motion out of Pika at times. I beat my head against the wall trying to get it to generate a car moving left to right in front of the camera, but never succeeded. I'm eager to try the "motion tracking" phrase you mentioned along with your other tips.
Any software that can get the final frame would work fine. ffmpeg is certainly an option.
Pika labs does seem to be updating fairly frequently so I do think they will allow you to generate longer motions soon.
I hope you have better luck with the motion, it seems easy to get more movement with the newer updates, but it is hard to control.
Is that with or without the new motion controls they announced on their Discord today?@@taoprompts
these are great and informative, thanks mate.
Thanks, and you're welcome!
very well! Thanks for all those tips
Your welcome!
fantastic work sir. really helpful.
Great sharing!
awesome tao
Pretty cool trick... Thank you dude.
You're welcome!
Nice!😍
I have a problem that the lightning doesn't align, it's a slight change, these are my parameters
visible eye blink, visible chest breathing, visible hair blowing, - camera wiggle, -neg morphing, erratic fluctuation in motion, noisy, bad quality, distorted, poorly drawn, blurry, grainy, low resolution, oversaturated, lack of detail, inconsistent lighting
What should I do?
What do you mean by the lighting doesn't align, are you using a reference image?
I see in your prompt you have "-camera wiggle" , for the -camera parameter it can only do zoom {in, out}, pan {left, right, up, down} or in a combination of them, and rotate {clockwise, counterclockwise}.
Super !!!
Thanks Tao
You're welcome!
Thanks for watching! Here's the UPDATED method to get cinematic camera motions in Pika Labs: th-cam.com/video/mnZfOvdUS5w/w-d-xo.html
Awesome tutorial, thanks. You'll have to give Augie a shot sometime :)
awesome and best explanation . Tao. love from pakistan❤
Thank you, Hello to Pakistan!
Hello, I have two questions to ask:
1. Capturing the last frame of the video may result in a color difference between the second video and the first one, as the captured image resolution is already compressed twice.
2. When merging the first and second videos, there are always inconsistencies before and after, such as pauses or inconsistent content
Do you have a good solution?
Hey, for your questions unfortunately Pika labs is still pretty limited when it comes to extending the length. This video shows a potential workaround, but you'll still run into issues as far as inconsistencies between the first and second videos.
I'll post an update video if they extend the video generation length though.
HI. What is the command prompt to do a dolly forward, dolly in . Not zoom but car sign in?
I don't think the dolly camera motion works atm unfortunately. I haven't seen any Ai video generators be able to get that specific motion.
i tried this method but the generated video is not consistent in color and contracts and even has different detailed for example the eyes and facial features. i had quite a few iterations but the result is not desirable. how should improve this?
It's hard to get this to work perfectly, Pika Labs is still limited in generating high quality long videos unfortunately. Perhaps changing the prompt to be a bit more descriptive in the details you want and using a high guidance scale parameter (like "-gs 20") could help, but there will still be some warping compared to exactly what you're original reference image was.
Thank you so much you solved my making pikalabs longer videos.. would you also tekl me which text to image would be good for creating mlre refrences images which should be also free .. there are two i knkw. First is Ideogram and bing dalle 3.. which one would you suggest.. i know you might be tell Leonardo
Ideogram and Dalle3 is a good start. I think Adobe Firefly is free atm which is worth looking into, I know they just updated their models.
If you are more technically inclined you can always use Stable Diffusion which I think is the best for free users, because of all the open source development being done and no censorship, although it has a high learning curve.
@@taoprompts thnx for the deep reply.. by the way.. I find Ideogram always works well this is the only which is free adm no limitation. Daale has limitations.. and everything else is not free
@@QuantumJumperTv Hopefully DallE will add some more parameters soon, even some simple ones like aspect ratio would help a lot.
Some very important points about there about motion tracking... Do you have some kind of document like a PDF that has also these terms I could use for my projects?
Thank you, I haven't created a pdf for this video, I'll keep that in mind for future videos though.
nice its was very helpful
Final Frame seems to be down. do you have any alternative sites to extract the last frame in a video?
Here's a video to jpg converter I use: ezgif.com/video-to-jpg
It will extract all the frames from the video instead of just the last frame though.
@@taoprompts fantastic! Thanks for the quick response!
I’m a musician and I tried but it says a waiting list, is there any way I get access to the old version on discord? 😢❤
Underneath the waitlist there is a set of links that says "Twitter Discord About us Careers".
Click on that discord link and you should have access to the old beta version.
Дякую, ти молодець!
How do i create my own thread? Doesn’t seem to work
Go check my newest video "Pika Labs Huge Update", in the timestamps towards the end I explain how to create a private message.
Yea, some images you can't make it to react. Even you use these action words flying,jumping, running. Nice tips
Thank you, the ai's still got a lot of room for improvement.
Thanks!
+1 doesn't work for some reason. is it a paid option?
It's definitely not a paid option, you should be able to see a pop up menu above the direct message when you click it.
Thank you blessing to you
Blessings to you too!
Final frame seems to be now a cheap version of pika labs and it doesnt extract the final frame anymore....
If you have issues getting video frames you can use this link instead: ezgif.com/video-to-jpg
Is it possible to remove the pika labs watermark?
As far as I know, it's possible to generate videos without watermarks in a special discord channel, but you need to reach out to pika labs social media team or discord people to get access.
thanks
Bro does it have a watermark?
Check reply on the pika update video.
how to remove pika labs watermark?
In the "faq" channel on their discord, there's a section for how to get non-watermarked videos. You need to contact their social media team to get the info on how.
Excelent
Thank you!
❤
i subscribe you, nice video bro ❤
pls make a prompt or animation landscape, like rain, waterfall, ship in the sea, sea ocean, or mountain. Thank you 😊
Thanks for subscribing!
Just too fast in the explanation
I'll try to slow down in future vids.
I think the speed is fine. Just pause the video if you need to
Super useful, thank you!