Dreamina Omnihuman Lipsync Tips From Real Use

• Omnihuman by Dreamina animates from just one image and voice input
• Works well with humans animals robots even TV faces
• Adding blur helps trigger movement like hair or skirts
• Slightly open mouths boost animal lip sync detection
• Background animation is hit or miss but head movement is solid

Dreamina Song and Guitar

Animated single image into a 15 second video clip of a woman singing AND playing her guitar.

Ok so now that Omnihuman’s out for everyone here’s what I found playing around with it and making a bunch of lipsync videos. If you've never heard of Dreamina check it out in my tools directory here. I had plenty of voices and pics lying around and tools like Zonos ready to go which made it all faster.

Dreamina gives you some simple but solid voices out the gate seems like they’re from Elevenlabs. I can already tell which voice is which just by hearing them.

Anyway this thing takes one pic and whips up a full animation (well full enough by current AI standards) of your character talking your text.

What kinds of characters work?

Sometimes it messes up if you're trying animal characters but mostly it handles all kinds:

• humanoid animals
• just plain animals
• humanoid robots (as long as there’s a face kinda)
• the undead
• even a closeup of a face on a TV with no eyes just a mouth

It animated all those nearly perfect in my tests. Not every single try was a winner but not far off.

Stuff I noticed

There is an NSFW filter but it ain’t super strict on visuals, but quite annoying for audio. Can't even mention 'how babes' or 'sexy women' anywhere, Kling has no issues with those words.

If you're using your own voice file it takes longer probably cause it has to transcribe it to text first. Weird they don’t let you just paste in the script if you have it already but whatever.

It doesn’t stop at the face. It tries to move the whole body even sometimes the background. Mixed bag there. Like in one test it animated these huge butterflies around a girl’s head—super cool. But in another the AI ignored glowing swirls behind a dancer. It also once thought a floating skirt was stiff and didn’t try to make it flow.

Still sometimes it nails it. One time it animated a hand pointing a gun and it looked spot on. That tells me it’s reading both the image and the audio or script to figure out what’s going on and then acts on that. And that’s how it should work.

What about animal lipsync?

This is pretty huge. We finally got AI lipsync that works on animals. Most tools ignore animal mouths or just skip syncing them. This one tries and sometimes it pulls it off. Want to boost your chances? Based on my tests here’s what helps:

Make sure the mouth is a little open. AI seems to need a clue where the lips or mouth are especially with animals. Human characters get more slack—even just a drawn line for a mouth is enough. But for animals it helps if you show how the mouth opens from the start.

You might be thinking “yeah but won’t that ruin the sync like it used to in Hedra?” Nah. That’s old tech. I tested with a zombie woman who had her mouth wide open to begin with and Omnihuman handled it fine. It just closed the mouth first then did the lipsync like normal.

Extra lip syncing tricks I’ve picked up

Besides the usual “reprompt and see what happens” trick (yeah yeah we all do that) here’s a few better tips:

• Don’t reprompt more than a couple of times. If the AI already saw your image+text it’s likely gonna make the same call again unless you change the input. Try tweaking the image or audio a bit for different results.
• Keep faces close and well lit. The closer and clearer the face the more precise lip movement will be. Avoid low-res stuff unless you like wonky animation.
• Dreamina can sing. I've got a video showcasing that separately. More than that, it can animate your singer as well as them playing an instrument: guitar, violine. It's just truly wild.
• And here’s a cool one—if something like a skirt or hair isn’t moving like it should, try adding a blur to it in your image. That little trick tells the AI “hey this part’s probably moving” and it often reacts by animating it. I did that with hair too and it worked.

So yeah that’s my breakdown of this wild new tool. Character consistency post is next up and that one’s a ride trust me. Bookmark this and swing back soon.

To try Dreamina right now go to https://dreamina.capcut.com The lip sync function is also called AI Avatar Generator there. It's also free to use right now. Probably not for long. And likely will get very slow the more people catch a whiff of it. 

Last modified 10 April 2025 at 14:54