I made two more AI-animated videos since I sent out my last post about falling down the AI art rabbit hole.
They’re both silly. All of these are super silly. I’m not trying to tell stories here. The AI is stilted and weird looking. No matter how good you can get the image, once you’re animating it, it looks janky. But still, it’s remarkable and it will get better and better with time.
What I discovered working on these videos is that in D-ID you can upload your own audio file. Before, I was using the prefab D-ID voices which are all dogshit and then recording new voices (either myself or with ElevenLabs AI). Now, I can make the audio first and generate the animation using that audio, which skips having to do a dub, which is time-consuming and tedious.
I’m not an animator or a voice-actor (though I would like to be a voice actor!) but I can only imagine the potential time-saving if you could record all the audio and just have AI animate the lips moving. Even if everything else is animated by people, this could be a super useful tool.
This first video is an ‘anime conversation’. The first two voices are AI and I wasn’t aware of the audio upload feature yet, so I dubbed new voices in place of the D-ID dogshit voices. The third voice is mine and I recorded it and uploaded it and had D-ID animate using it. Interesting note: I tried several other anime faces for this character but the software would either not detect them or, in the case of one handsome fella, said it broke terms of service because he had a bare chest.1
These are YouTube Shorts so they’re designed to be viewed on a mobile device, which is why they’re formatted in this resolution. I should probably upload these to TikTok also.
AI Anime Conversation
The second video I went all-in on Western tropes. The music, the sound effect of the hawk screeching, the gunslingers. Midjourney for all the images in this. I recorded both voices and uploaded the audio to D-ID. This one was fun to make, but it’s a little sweary.
It’s all very neat and somewhat scary. I think about the one voice-acting class I took and how we had to do dubs and whatnot and it’s wild to do voices and be able to have them paired with an animation so quickly and easily. The future is coming and it’s wonderful and terrifying.
P.S. One thing I haven’t used in any of these is ChatGPT for the writing. I’m not writing any of this. I’m just speaking off the top of my head and making shit up as I go along.
Speaking of bare chests, Midjourney also prohibits any kind of nudity and bans lots of words from being used as prompts. However, I have accidentally generated a number of topless women when I’m using prompts for barbarian art. It seems the software will happily pop out naked breasts if you’re using the style of Frank Frazetta and Boris Vallejo…