The Princess' Butterflies - Using an AI Image to Video Generator to Animate Art That Started as a 23 Year Old Rough Pen Sketch

Back in October of 2023 I used Prome AI's Sketch Rendering feature to color a 23 year old rough pen sketch I did of a princess watching butterflies, sitting next to a waterfall in the forest. You can read more about that in my previous article, The Princess' Butterflies - Using an AI Art Generator to Color and Enhance a 23 Year Old Rough Pen Sketch

The three images I used to create my final image
(bottom right). Notice how closely the AI has followed
my line art sketch (top left) but in the first image
(top right) has misinterpreted the waterfall as some kind
of large tree stump.

Prome's render of the artwork was pretty good but I did end up creating a composite image myself from two of the best renders, to get my final image.

Since then AI Image to Video has come a long way, particularly when it comes to creating animation from artwork and maintaining the art style for the whole generation.

AI still does much better with turning photographic images to video but I found an AI site, PixVerse, that seems more focused on animated styles. Initially I used the site to animate my cartoon TET avatar, which you can read about in my Animation and Video life blog.

Since PixVerse did such a good job at maintaining my art style, by not actually applying any of their art style models to the image (which are all very distinct animation styles like Anime), I thought it might be worth a try running my Princess AI image through to see what happened.

Overall I was quite impressed. I generated four different videos of nine seconds each, getting something reasonably coherent each time.

I did have a few issues like butterflies morphing into a single butterfly, butterflies randomly appearing out of nowhere, and the usual deformed hands (but only if you really look because it tended to happen if the hand moved fast into position).

PixVerse created this 'happy accident' when it randomly
generated a magical circle after misinterpreting my prompt.

One generation I got a happy accident where I attempted to get the butterflies to fly in a more circular formation. Instead the AI interpreted the word circle by adding a magical light circle (kind of like a spell had been cast) appearing around the butterflies.

I decided to take my four generations and edit the best bits together into a single video clip, adding sound effects, music, and even the princess whispering some dialogue. Which was an after thought. It just seemed odd to have all the sound design and the princess, who is clearly speaking, not having any audio.

Although PixVerse does have a lip sync feature that I could have used, I ran out of paid credits, and didn't want to purchase more because I didn't have any immediate need for them after this project - which, in itself, is just a test project. So I just tried to come up with some dialogue using ElevenLabs AI voices that approximated with her lip movement.

You can also generate sound, speech, and even extend the video right in PixVerse but again, I didn't want to purchase the extra credits on a test animation.

The resulting video is below. I tried to edit it so that if you loop it, it's not too jarring, though obviously the dialogue becomes a bit redundant after the second play through. I hope you like it. Feel free to leave a comment to let me know what you think.