Five AI Generative Image to Video Tools For Animation You Can Try Free Right Now

The Emo Girl Character created by Start Animating.

AI generative video isn't new but it is the next big thing in the visual imaging space as various development teams work to perfect the generated output. 

Just like generating still images AI video sometimes struggles with physics, arm and hand movement, and the general structure of things.

However it is getting better and, as is the catch cry of all AI development, this is the worst it will ever be, because it's improving fast.

If you're an animator one current potential use of generative AI video is to animate your key frames, as opposed to generating something entirely from a text prompt. Starting with an image helps to keep your characters and art style consistent across AI generations.

With that in mind I tried five, free image to video AI generators to see what their potential might be and whether they can handle cartoon style characters well. Note all but the last entry on the list do not create any sound with the video output.

Toon Crafter

Toon Crafter is the generative AI video everyone is currently talking about. It lets you enter a sequence of images (key frames) and it'll do a pretty good job of animating the tweens.

This is Toon Crafter's free playground interface. I was never able to generate anything with it. Underwhelming given all the hype.

I tried for days to get something to generate on the free playground with no success. It would just sit there ticking over for hours on end, to the point where I could've done the animation myself faster with Cartoon Animator.

It's supposedly good but I'm not paying for something I can't even trial on my own artwork.

Leonardo Image2Motion

At this point Leonardo's Image2Motion is very much a crap shoot. You can only input images from your own image generations (or ones from the public feed). There is a hack to upload an image but I actually forgot how I did it (maybe through the Canvas Editor I think?). 

You have one control. A slider for how much motion you want to add. Then you generate and get what you get. I've found it works best on images with a distinct foreground and background for which it really excels at doing cinematic panning shots.

I didn't have the option to prompt the action in either of these images.
Amee, on the left, uses a low motion setting which I think is giving
her gastric reflux, while the pan around the alien on the right is
the kind of shot Leonardo's Image2Motion seems to favor.

Krea AI Video Generation

Krea's AI Video Generation is not unlike Toon Crafter in that you can add in a whole series of key frames but also, you can add in different prompts for different stages of the generation too.

I've seen some fairly good results when it comes to morphing one image to another but using it to draw tweens between similar key frames is a little more tricky to get usable results. Also the 'Render' setting produced images closer to my cartoon image key frames than the 'Animation' setting did.

Of all the Generative AI's listed this one has the most options for influencing the output.

Krea's User Interface has the most potential for fine control of the output. Unfortunately
none of the style settings would maintain the art style of my key frames.

Luma Dream Machine

With a very simple interface, where you can either enter a prompt describing what you want to see, or you can start with an image and prompt for what you want to see happen within that image, Luma Dream Machine is still quite impressive.

On the free tier you currently get five generations in any given day, and 30 generations per month total, no commercial use. Which is seriously frustrating and I hope they'll extend that daily limit again once they're better able to meet the demand.

The video output is very good but you still get some weird things happening with arms and hands if you look closely. Each generation is only five seconds. There is this nice feature where you can automatically grab the last frame from any generation and extend the clip (and the AI will automatically stitch it to the previous clip as well).

On the surface output from Luma Dream Machine looks exceptional. My original
still images are on the left. In both cases the character was prompted to walk to the desk.
In both she walks right past it. Despite that, the background is accurately extended
in the same art style. The character subtly changes as she moves across the room,
with questionable arm movements. (Also I did not prompt for that mysterious person
who appears in the bottom frame). 

I was quite impressed with the output though I think there's probably a learning curve for how you prompt the action. My first prompts were quite simple and I think the Enhanced Prompt check box, that I had ticked, tended to add more action than I actually wanted.

Ultimately the results are the best of those listed so far just because it stayed closest to the original look of my key frames. The down side was the AI didn't seem to follow my prompts very closely and introduced elements not even mentioned in the prompt.

Hedra

Hedra is specifically for creating animated avatars that talk in a natural, realistic way. More than just lip syncing it animates the avatar's entire head and body (to some degree so the body isn't just still).

You can actually prompt an avatar but I suspect more people will use the option to upload an image. 

There's no way to crop the image yourself. Hedra looks for a face in the image and then crops accordingly. How much body you get depends on the image but Hedra favors a fairly close crop of just the head.

Hedra clearly specializes in making photographic avatars speak. When I tried it with a very cartoony 2D image, while it recognized the face, the output was clearly using a real human model for the eyes, mouth, and face structure. The whole thing looked creepy and unusable.

However, if you have something photographic or very close to it, like an anime character rendered with a slight 3D style, the results are extremely convincing.

Hedra's User Interface is very simple but produces the best results of all the services listed
provided your avatar is photographic or in a semi realistic style. You can see an example
of this character animated (from a different photo) on my Art Time Productions Facebook page.

Currently Hedra's TOS states that you can only upload images of yourself, though they have no controls for policing that, and I'm guessing uploading digital avatars of non-existent people is fine (hopefully, because I see Hedra being used by AI influencers a lot).

Small Rant Section About Generative AI

I've heard a lot of nonsense about Generative AI which I don't intend to vent about here. The fact is Generative AI is causing a shake up of all creative services. It's here to stay. 

As much as some creators think it's going to take their jobs, it's also creating new jobs, that right now, are not hard to become an authority in, if you already have years of creative experience and output behind you.

Now is the time to explore all the opportunities for increased workflow and broader creative scope that this technology brings. 

Eventually the field will become too complex to break into easily. Learning while even the people developing the technology are learning is better than letting it all pass you by out of some misguided propaganda that ALL AI is 'stealing' from real artists. 

Non creatives in the commercial sector don't actually care how their art is made. This is actually your chance to stay informed, and become a curator of the best AI generative tools, learn how to use them better than non creatives.

It's no different to photography. In a day and age where almost everyone has a camera on their person, and we have stock image libraries for days. There is still a need for professional photographers and video camera operators. AI is not coming for you, nor will it make you redundant (and, if you didn't notice, a lot of the tools in this list aren't generating any 'stolen' images or styles).

AI is just shaking up how humans create. It's not bad. It's just new. You'll always have the option to charge a premium for art created 100% by a real human, if you really do wish to not engage.

o---o--- ---o--- o---

Did you find this article useful?
Subscribe to my newsletter and get the
latest articles delivered to your inbox.