Image-to-video comparison: Pika vs Runway (August 2023 - newer version in the description!) 

4 Months only and sora ai dropped out lol

Thank you! Good call on the frame rate and resolution differences. I may have to check out Topaz, that's the third time I've heard about it this week. I'm happy to wait for Pika and Fulljourney to add upscaling because when Gen-2 "improved" it took a lot of the fun out of it.

@@aivideoschool Can't say enough good things about Topaz. Besides the uprez and interpolate it also will give you REALLY smooth slow motion. Very useful extending the length of clips.

You can now manually set your frame rate with Pika Labs using the -fps parameter.

@@tyberwhite Thanks! I just tried it. Butter smooth!

wow, -fps 24 makes everything so much smoother! thanks for the heads up.

My pleasure!

Pika labs free or runway free. Answer me

I just saw FullJourney allows LoRAs now (haven't tried it yet). I think Loras will help with that consistency even more. Being able to create a custom face you can prompt with will be huge.

Thanks for replying, I think very soon advanced release will cater consistent characters, facial, expression, etc@@aivideoschool

Totally agree. Someone else pointed out in the comments here, Pika now has a "-fps 24" control you can add to the end of the prompt. So much smoother! Runway's tools are great, and their non-generative AI tools (Magic Tools) are wildly underrated.

I'm so glad it was helpful!

There will be great economic problems as AI evolves and jobs are lost in al sectors. We've already seen bank tellers, Copywriters and marketing being first, graphic designers and art directors next. Filmmakers, music composers, news anchors, draftsmen, architects, fashion designers, taxi and delivery drivers, etc. As robotics and AI evolve, there will be fewer and fewer jobs that are not cheaper to do than with technology. We will lose the ability to think and create as we rely on technology more. We will have more people without occupations that still need to be fed and housed and kept busy so they don't revolt. This leads to a dystopian society of walled off communities of elites and ghettos of the masses. If society wants to survive, then there needs to be legislation to control the extent of AI. We will have media completely made up to create narratives that the ruling class figure will control the masses and keep them in their place. We are already seeing the chaos and societal breakdown occurring as media is being controlled. When it becomes manufactured, then the dystopian future we see in sci-fi movies will be reality.

@@tonyr.4778 That was deep. Did you generate this text with ChatGPT?

​@@tonyr.4778well not everything is about money, and for now, even with AI human factor is still a major one. It's like people were panicking when photography was invented. Just chill. It's only a tool.

Someone on the Pika Discord mentioned starting with a landscape and adding things to it via text prompt. I could mostly get that to work by telling it where to place the new thing "a tank/nomad/horse goes by in the foreground" which you can see at 05:12.

also, out of curiosity, is your own artwork landscape or portrait mode? Pika does both, Runway only seems to do landscape.

@@aivideoschool Cool! Never thought of doing it like that. I have had a subject in my image, and tried animating that subject via text. That doesn't seem to work

@@aivideoschool Never thought about that, however, my drawings were 1920x1080, so it should have worked. Seems like cartoon drawings are harder to animate in RunwayML, unless they depict humanoid subjects, while Pika on the other hand can easily handle things like cartoon ducks etc

Thank you! That's what I aim for.

I'm glad you enjoyed it. Thank you for watching!

Thank you! I just went back and looked and I wasn't using the -motion parameter on those. (That might have been a choice to use "default mode" for both.) I'm loving the new -fps parameter in Pika. I've noticed when using -fps 24 and -motion 4 it sometimes creates too much noise/unexpected motion. Not always but sometimes.

@@aivideoschoolWhat are you doing with the -fps parameter? I haven't played with that one. Just the -motion and -ns parameters.

Thank you!

Optimistic to think that this will only be used in storyboarding. At this rate in a couple of years, actors wont be needed, camera men, directors etc. So let's chill and take it more slowly

I was dragging my feet on Kaiber because they were one of the first to go from free to pay-for-credits, but everyone is going paid now. Kaiber seems focused on creators/creatives (rather than corporate/enterprise) which I love so I'm checking them out now.

@@aivideoschoolJust paid only? Or Free tier plus paid tier like Leonardo AI and others. Pika Labs is still free, free only in fact.

Thanks, I don't think Pika had negative prompts when I made this but I am very tempted to go back and try now to get her hair to stop blowing :)

I feel like image to video actually gives better results with more motion than text to video.

It's a weird trade off: "you don't want it to be blurry? Lower the motion so you get an almost static shot." The original Gen-2 model had a lot more motion and none of the blurriness but there were a lot of hallucinations. I think that's what all the models are over compensating for.

I like Pika too!

Ixi lá ele

I've been having great fun and quality using Leonardo AI's new Photorealistic gen mode and taking the image generated and using it with Pika Labs "/animate" command. I believe the output is on par with RunwayML at that point. @aivideoschool

Pika is free. Also, check out FullJourney, which is free and based on Zeroscope. Here's a tutorial ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-eaV9wzRDTMQ.html

I tried to use pika, register on discord and I couldn't use it instead am still on waitlist

Runway just released Motion Brush. That might be helpful, combined with a prompt. How to use Runway's Motion Brush ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-soZbh9O25ig.html

You can do a lot of AI stuff on Discord (Morph Studio, FullJourney) or websites (Runway, Pika Labs, Leonardo AI). All of that can be done on a Chromebook or phone. It's only when you want your own computer/GPU to generate things locally that you'll need a powerful computer.

Funny you say that... when I recorded this video Motion Brush didn't exist, but once it was released, I went back to test that exact bus ride clip you mentioned and it turned out great. I love Motion Brush. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-soZbh9O25ig.html

@@aivideoschool Oh thanks. im watching it. i see the proximity should be -10 for one object movement- that was bugging me all this time.

U helped me a lot. Im doing animated nfts of ninja ducks rowing in boats, i couldnt get the boat and waves and clouds to all move but not their faces. Even with magic tool. But i never tried -10proximity.

Nice! Those shots look awesome.

Really great quality! What do you think of Leonardo AI being fed into Pika Labs (/animate)? The two green female android clips in this video were made this way: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-IPLFHNcQY-c.html

just checked out your creations on your channel and they do amazing!! I'm still debating on which platform to subscribe too lol

I know many people go that route for image generation but in my experience, doing a short, concise prompt with a specific instruction works better for image-to-video. I'm genuinely curious if anyone has good tips involving longer i2v prompts.

In Gen-2 there's a settings button when you generate (currently in the bottom left). If you open it, you can Remove Watermark. If you've already generated the video, try Runway's own Inpainting tool. It might work?

I know, I've had success and misses with both. One tip I heard that I found helpful was remembering the AI can see the picture too so I might say "a fish swims out of the hole in the ship" or whatever you're going for. And then keeping the action limited to one specific thing that can be shown in 3 seconds. It's still not perfect but then I fine tune from there or hit the refresh button. Good luck!

Runway updated their model recently and it looks more realistic than when I recorded this, but I totally agree. I hate the cartoonish AI look that isn't quite animated but isn't real looking either.

@@aivideoschool It's impressive, but still a long way off from me using either option.

Nice! Pika now has 24fps if you add this to your prompts: -fps 24

Runway's AI rotoscope/green screen tool is pretty impressive

With a paid plan you can

@@aivideoschool thanks bro🙏🏻

Yeah, it would be great if they had the "extend" option like RunwayML. But you could always use a video editor, or automate the process using ffmpeg, grab the last frame, and resubmit it to Pika Labs with the "/animate" command with a text prompt telling it what to do next with that final frame.

Thanks, I agree. Runway had an early lead but the controls and quality from Pika have surpassed them in the past few months. Hopefully the competition means higher quality video and more creative options.

No offense taken. I just want to clarify that the quality I referred to had to do with technical aspects like HD resolution and frame rate. 3 months ago when this was recorded, Pika was only like 12 fps and half the resolution that it is today. Pika has surpassed Runway and it's still free

@@aivideoschool thanks for the response! I understand your point totally and I'm glad you didn't interpret my comment in a bad way.

FYI - I don't do promos for anyone, I just share what I find interesting and am enthusiastic about. If you watch the whole video, I say they're each better at different things.