Loading video...
Video Failed to Load
I'm putting MiniMax H3 and Seedance 2.0 head-to-head in this lip sync test using Pollo AI Using the exact same prompt and audio sample. I will leave it up to you to decide which one you think did better. A lot of the time it comes down to preference... show more
19,879 views • 1 month ago •via X (Twitter)
33 Comments

Prompt: "1970s sci-fi fantasy style with a practical-effects feel. A space cowboy from a distant planet rides a strange alien beast up a rocky hill toward a mountain summit. The world should feel like an old retro sci-fi film with handmade creature effects, theatrical terrain, painted skies, and a slightly camp but cinematic tone. Lip-sync note: Use the provided audio sample as the master timing reference. The character must lip-sync precisely to the audio. Keep his face and mouth clearly visible at all times. Prioritize tight close-ups and medium close-ups so the lip sync can be judged easily. Shot 1: 0–4s Tight close-up of the space cowboy riding, with the strange landscape moving behind him. Clear view of his face, expression, and mouth as he performs. Shot 2: 4–8s Medium close-up from the side/front angle as he continues up the hill on the beast. The motion feels adventurous but the framing stays tight enough to clearly see his lip sync. Shot 3: 8–12s Close-up at the summit with the tiny distant city visible behind him. He finishes the performance with a proud, cheeky expression while looking out, then slightly toward camera." I also used these two images and a 12 second audio sample

For more information about Pollo ai:

@itsPolloAI H3 holds its own with physics, lip sync, human exprssion and maintaining the world as it cuts. Crazy.

@itsPolloAI Yeah I've been quite impressed with H3 considering it's cost and default res of 2k as well.

@itsPolloAI H3 is sticking around for a while 😉

@itsPolloAI Its so close! I like both! great work

@itsPolloAI Thanks mate me too

@itsPolloAI H3 did great on the lip sync, it still smooths out the style in my opinion. I feel like the Seedance clip has more of a crisp style.

@itsPolloAI Yeah each one has a different tone to it.

@itsPolloAI Minimax H3 cooked🔥🔥🔥

@itsPolloAI We have the power now mate hehe

@itsPolloAI Lip sync quality is evolving fast and competition keeps pushing everyone forward

@itsPolloAI Agreed it's a great time to create

@itsPolloAI Reminds me of Marshal BraveStarr

@itsPolloAI Classic toon

@itsPolloAI It seems like seedance won.

@itsPolloAI Same here Damn you @CaptainHaHaa I can't choose !!!

@itsPolloAI hAHah yeah me too sometimes I think it comes down to purpose in the end.

@itsPolloAI Great video Captain! But I miss universe and your uniform. 😅

@itsPolloAI Not to fear mate, I've been thinking about them too. I have a few things in the wings just need to balance my time.

@itsPolloAI Minimax wins for being an opensource chad!

@itsPolloAI HAhah

@itsPolloAI Same prompt and audio sample is the fairest lip sync test.

@itsPolloAI I loved the look of Pollo when I played with it.. and there model isn’t bad itself.

@itsPolloAI Yeah I've been checking it out, its pretty good.

@itsPolloAI tomato, tomatoe. on cost value alone H3 takes it.

@itsPolloAI HAha pretty much, there are some subtle differences but the cost just nails it.

@itsPolloAI They are both excellent. I've been so impressed with H3, given the lower cost. But, compared to lipsync options a year ago, both are a tremendous step forward.

@itsPolloAI Yeah we have a few options now we is very nice to see. Ultimately its a big win for creators

@itsPolloAI Same prompt, different results

@itsPolloAI Fight scenes often trade fluid movement for recognizable characters. Atlas Cloud could measure that balance by checking MiniMax H3 and Seedance at contact frames, occlusions, fast turns, and the quiet moment after impact.

@itsPolloAI One is free to use the other one you have to pay to use. Which one would you choose if you don't have money to throw around.

@itsPolloAI Head-to-head lip sync comparisons are always fascinating to watch. Lip sync accuracy and natural head motion really set the top models apart. We've been running photo to video lip sync pipelines over at if you want to test another setup.
