For a while now, I’ve been wondering how AI would interpret my old illustrations that I like, and what kind of concept it would present if I converted them into videos without adding too much detail.
I then seized the opportunity, rolled up my sleeves, and immediately turned to the Google search engine. I conducted my search using phrases like “free AI video.” Previously, when I tried it, at least the Hailuoai site offered quality demos in its free version, albeit with some limitations. Now, however, the site, which was only in Chinese, didn’t even allow any production with a watermark.
I’ve noticed that many AI sites advertised as free now only use this as a phishing tactic until you sign up with your Gmail address. I haven’t been able to get any testing work from them either.
Finally, I looked at Gemini’s plans. I chose the one on the far left, but I paid 200 TL instead of 49 TL, probably because of another membership.
This membership only allows us to produce two videos per day. This was slightly disappointing, as it’s far too little, even for testing purposes. But the results it produced weren’t bad.
As I mentioned, I was curious how the AI would perceive my drawings as they are, and I didn’t enter any explanatory prompt. I just told it to convert this illustration into a video.
Gemini creates the video along with the audio. You can enter separate prompts for the video and the audio.
This illustration, which I created in 2021 for the theme “Sepia,” was used as the cover image for a magazine published by the Irish Times. Samsung also used it as a screenshot for their tablet launch.
Without adding a prompt, it gave a nice result that was close to the concept. And then I added this drawing:
I provided a prompt to rotate the drawing according to the measurements, ensuring it adhered to the specified dimensions. However, despite the drawing not explicitly showing breasts, it was interpreted as sexually suggestive. It appears there’s a highly sensitive filtering system in place.
Thinking they would process it quickly, I made the following desperate request:
: ‘) But it interpreted this prompt for the first drawing and created this video:
Despite the great result, either the AI genuinely misunderstood, or, to put it mildly, I was brushed off by it. 🙂
If I were to offer a general observation, I’ve noticed that these services are no longer offered with such generous credits as in the early days, but with more controlled credits. Even with the cheapest package, when I paid, it only allowed me to produce two short videos per day. I will add an update to this post once the process is complete, detailing the total number of videos I was able to produce.
I will use my remaining credits for other illustrations and update the post accordingly.
Update:
Up to this point, I’ve noticed he understands the emotion in the drawings, but in terms of visual style, it tends to create more concrete objects. It tries to adhere more to the meaning of the reference you provide. If it perceives a wolf in a suit, like in the example, it immediately puts the wolf in a suit there; he doesn’t get hung up on abstract or stylistic effects.
Sometimes, it seems like it takes a shortcut by using a previously created video to save energy. I didn’t like it wasting my rights this way. I’m not sure if we have the right to request a preview before he produces this; if anyone knows, I’d appreciate it if they could let me know.
Thank you for your time. If there’s anything you’d like to correct or add, please leave a comment below.
Semih Akgül