Alpha Mom

New body-physics test for the best AI video tools: The newly released Grok Imagine 1.5, the king Seedance 2.0, Google’s “revolutionary super-duper” Gemini Omni Flash, and the old-timer Kling 3.0 Pro. This time I tested how each model handles realistic running motion, body movement, fabric physics, and natural secondary motion. And the result was much less obvious than I expected. Believe it or not, I can’t call Seedance the clear winner this time. Each model got several attempts. - Grok Imagine 1.5: The new version is finally available on the official Grok website, so I could properly test it. Honestly, I didn’t notice a massive improvement over the previous version. Still, the result was acceptable. It produced the most cartoonish image, and the woman runs as if she’s wearing heels, but the body physics were decent enough. It also understood the instructions quickly and followed them correctly. - Kling 3.0 Pro: The old man decided to test my patience. It repeatedly blocked a simple running scene as adult content, then misunderstood the instructions several times. The successful result has the most realistic lighting and frame rate, but the actual body physics look strange. It almost feels like loose foam padding is bouncing inside the leggings. There are also several visible artifacts and unnatural movements. - Gemini Omni Flash: As usual, it gave me that strange slow-motion, low-FPS look that Google models seem to love. But it didn’t censor anything, understood the instructions immediately, and produced a beautiful, realistic result. Surprisingly, this is the output I liked the most in this test. - Seedance 2.0: Seedance also blocked a couple of generations, just like Kling, but eventually produced a strong result. It delivered the most beautiful and visually appealing footage, but I honestly expected better physics. The video looks great, yet I can’t confidently call it the most physically accurate result. - My ranking: 1. Gemini Omni Flash — not perfect, but the best overall result for me 2. Seedance 2.0 — visually stunning, but I preferred Omni’s physics 3. Grok Imagine 1.5 — cartoonish look and strange running, but still acceptable 4. Kling 3.0 Pro — the longest wait, the strangest physics, and the most inconsistent result Do you agree with my ranking? #AIVideo

@ZFmepiU0Q468384 这次确实没能翻身，虽然比上一次好一点，但离 Seedance 和 Omni 还有明显差距 😅

@YourAlphaMom 毫无悬念，Grok又在动作和画质上输了😭

Another body-physics test for the best AI video models. This time, Elon’s pride and joy Grok Imagine 1.5 faces the old but stubborn Kling 3.0 Pro, China’s flagship Seedance 2.0, and America’s shiny newcomer Gemini Omni Flash, which developers called “revolutionary,” although I’m still waiting to see the revolution. The new battlefield: jump-rope exercises filmed from the front. There is a lot to analyze here. But let’s be honest. We all know exactly where everyone will be looking. Or rather, at which two points. The result was surprisingly close. - Kling 3.0 Pro: In this test, Kling was the only model that detected adult content in a completely harmless fitness scene and repeatedly refused to generate it. Which is especially funny considering how often Kling is used to bypass censorship with celebrity faces, yet apparently a woman jumping rope is where it draws the moral line. After several attempts, it finally produced a usable result. The lighting and overall realism are strong, as usual, but the actual body physics are the weakest here. It also missed the visual instructions from the prompt, giving me an older-looking model with less flattering proportions instead of the athletic, curvy character I requested. Not terrible, but definitely not Kling’s finest performance. - Grok Imagine 1.5: Apparently, Elon’s creation performs better from the front than from behind. Grok understood the task immediately, generated quickly, and didn’t complain about censorship. The image still has that unmistakably cartoonish Grok look, but the jumping motion is surprisingly fun and the body physics are actually decent. It feels more like a cutscene from a video game than real footage, but this is still one of Grok’s better results in my recent tests. - Gemini Omni Flash: Once again, Omni produced a beautiful and polished image. Google’s usual slow-motion, low-FPS effect is normally one of its biggest weaknesses, but in this specific test it actually worked in its favor. The slower movement makes the secondary motion much easier to see, and the body physics look smooth, convincing, and surprisingly natural. I liked this result a lot. - Seedance 2.0: Seedance performed much better than in the previous body-physics test. The footage is dynamic, vibrant, realistic, and visually the strongest of the four. The physics still don’t look completely natural to me, but the overall result is so convincing that it deserves first place, or at least a shared victory with Omni Flash. Omni may have slightly better body physics, while Seedance wins in movement, realism, energy, and overall image quality. - My ranking: 1. Seedance 2.0 and Gemini Omni Flash. A shared first place. Omni wins on body physics, while Seedance wins on overall realism, visual quality, and dynamic motion. If I absolutely had to choose one, Seedance would take it by a very small margin. 2. Grok Imagine 1.5. Still behind the two flagships, but noticeably better than in the previous tests. 3. Kling 3.0 Pro. Too much censorship, weaker instruction following, and the least convincing physics. The realistic lighting and natural-looking footage save it from being a complete failure. What’s your ranking? And if anyone wants the prompt, ask in the comments and I’ll share it. #AIVideo

@FilmFireTV That sounds interesting! I’d be happy to take a look when early access opens. Thanks for thinking of me 😊

@YourAlphaMom Since you're into testing and creating we'd love for you to join us when early access opens. We want to see more! 😃

Kling 3.0 vs Gemini Omni Flash vs Grok Imagine vs Seedance 2.0. A new battle for the best AI video generator. This time I tested the leading models on something different. Not a chase. Not a fight. Not a crazy action scene. A beach workout scene. Sounds easier, right? Not really. I made the prompt harder by asking each model to show one character doing three different exercises in the same video: jumping jacks, squats, and a cat-cow stretch. At the same time, the video had to look stylish, beautiful, realistic, and good enough for a top Instagram fitness account. Each model got around 4–6 attempts. There were also some safety rejections. Seedance surprised me the most here. I only managed to generate 2 usable videos with it because the other attempts were rejected by the system, even though the prompt was just a beach fitness scene. Kling and Gemini Omni Flash also had some rejections, but fewer. Grok had no safety rejections in this test. My take: - Kling 3.0 did better than I expected. You can still feel that it’s getting outdated, but the body looked reasonably realistic. It didn’t really give me that premium Instagram-fitness look from the prompt, and the exercise physics were not perfect, but honestly, this was a decent result for Kling. - Gemini Omni Flash was very strong here. The instructions were followed well, the exercises were actually there, the video looked expensive, and the realism was solid. This may be the best Gemini Omni Flash result I’ve personally had so far, and in this specific test, I think it was the winner. - Grok Imagine was disappointing again. I did not use the new Grok Imagine 1.5 here, because it seems to be available through the API only, and not in the regular tools I currently use. So as a regular user, I used the official Grok website, where an older version is most likely still running. Visually, the result was acceptable, but the prompt was basically ignored. Instead of a workout, it turned into some kind of dance, and at the end the girl suddenly did a split and surprised all of us. - Seedance 2.0 was strong as always, but I expected even more. The instructions were mostly followed, the physics were good, and it had the usual Seedance energy, motion, and FPS advantage. But in this scene, extreme speed and intensity were not really the main point of the prompt. Maybe the result could have been even better if the system had not rejected so many generations and I had more outputs to compare. My ranking for this specific scene: 1. Gemini Omni Flash — beautiful result, strong instruction following, and probably the best Omni Flash test I’ve done so far. 2. Seedance 2.0 — very good, and I’m sure many people will still put it first. But for this scene, I think Gemini Omni Flash did better. 3. Kling 3.0 — decent result, better than expected. 4. Grok Imagine — weak result in this test. I hope the new Grok Imagine 1.5 will be much better once I can test it properly through my tools or the official website. Enjoy the test, and let me know your ranking. #AIVideo

@FilmFireTV It picked the right test to finally show up 😄

@YourAlphaMom Omni Flash is puttin' in work!

No worries, and that context makes your point much clearer. For preserving original creature designs across shots, Seedance does sound like the stronger tool right now. My tests are narrower benchmarks focused on motion and physics, so they won’t cover every production workflow. I’m planning a dedicated same-reference-image comparison too, which should be much more relevant to this use case.

Sorry, I may have sounded a bit harsh earlier and caused some misunderstanding. The images I uploaded are all original alien creatures designed by myself, but the two models generated results that looked completely unrecognizable. Seedance, on the other hand, replicated my designs almost perfectly, which makes it possible to achieve an early prototype of industrialized AI filmmaking. But for original and abstract subjects, the current technical level of those two models is simply not usable yet. So that’s why I feel your comparisons lose their meaning for creators like me.

New body-physics test for the best AI video tools: The newly released Grok Imagine 1.5, the king Seedance 2.0, Google’s “revolutionary super-duper” Gemini Omni Flash, and the old-timer Kling 3.0 Pro. This time I tested how each model handles realistic running motion, body movement, fabric physics, and natural secondary motion. And the result was much less obvious than I expected. Believe it or not, I can’t call Seedance the clear winner this time. Each model got several attempts. - Grok Imagine 1.5: The new version is finally available on the official Grok website, so I could properly test it. Honestly, I didn’t notice a massive improvement over the previous version. Still, the result was acceptable. It produced the most cartoonish image, and the woman runs as if she’s wearing heels, but the body physics were decent enough. It also understood the instructions quickly and followed them correctly. - Kling 3.0 Pro: The old man decided to test my patience. It repeatedly blocked a simple running scene as adult content, then misunderstood the instructions several times. The successful result has the most realistic lighting and frame rate, but the actual body physics look strange. It almost feels like loose foam padding is bouncing inside the leggings. There are also several visible artifacts and unnatural movements. - Gemini Omni Flash: As usual, it gave me that strange slow-motion, low-FPS look that Google models seem to love. But it didn’t censor anything, understood the instructions immediately, and produced a beautiful, realistic result. Surprisingly, this is the output I liked the most in this test. - Seedance 2.0: Seedance also blocked a couple of generations, just like Kling, but eventually produced a strong result. It delivered the most beautiful and visually appealing footage, but I honestly expected better physics. The video looks great, yet I can’t confidently call it the most physically accurate result. - My ranking: 1. Gemini Omni Flash — not perfect, but the best overall result for me 2. Seedance 2.0 — visually stunning, but I preferred Omni’s physics 3. Grok Imagine 1.5 — cartoonish look and strange running, but still acceptable 4. Kling 3.0 Pro — the longest wait, the strangest physics, and the most inconsistent result Do you agree with my ranking? #AIVideo

@FilmFireTV Really appreciate the shoutout! I’m just putting these models through the tests their demo reels tend to avoid 😄

If you're learning the tools @YourAlphaMom can help. Creators are paying attention to details. #Kling #Gemini #Seedance #GrokImagine #FilmFire

@TsarBunza Then she needs to start asking Elon for royalties 😄

@YourAlphaMom This absolutely confirms it. Grok Imagine was trained on my sister

Calling the reviews meaningless because they don’t cover every possible workflow is a bit of a stretch. This test is specifically about motion and body physics, not a complete filmmaking pipeline. I haven’t explored Grok’s reference workflow enough to judge it properly, but Veo and Gemini Omni are far from unusable here. Both support multiple references. Veo already supports first and last frame control, while Omni supports the first frame and Google lists last-frame control as coming soon. You can also upload two reference images and describe the intended transition between them directly in the prompt. Multimodal reference is important, but it doesn’t invalidate a physics benchmark.

@YourAlphaMom All your reviews are meaningless. When it comes to multimodal reference, except for Seedance 2.0 and Kling, the other two are basically unusable. Without multimodal reference, how are you supposed to make original creature-character videos of your own?

@ZtohAic Exactly the kind of peer review this study needed 😌

nice bounce.

@FilmFireTV I take my scientific responsibilities very seriously 😂

@YourAlphaMom Good test! Very thorough research.

@techkasia Yes. They can generate explosions and cinematic action, but a normal running cycle is still the final boss 😅

@YourAlphaMom When AI will learn to do run 🤦🏼‍♂️ it’s always those missing or double steps, it’s so simple left, right, left, the same peace!

@neuraltitan Exactly. Explosions and fast cuts can hide a lot, but simple everyday motion leaves the model nowhere to cheat

@YourAlphaMom Ai mimicking human movemants not in action scenes , space scenes, high impact scenes but in everday occurance scenes is were most fail.

@AIZEN30XX Yes. Better than in the previous test, but nowhere near Omni or Seedance yet

@YourAlphaMom Dang Grok is really bad with physics

@TheMainKrispy @agnesai_sapiens I might include Agnes in a future comparison if I get a chance to test it properly

@YourAlphaMom You forgot about @agnesai_sapiens their video generation is pretty cool on their video v2.0 generation.

@htuna_ @xai That’s a good idea. I’m planning to run a same-reference-image test too — same image, same task, all models. We’ll see who really handles it best.

@YourAlphaMom Create videos using the same reference image for all of them and show the results. @xai Grok will be far ahead.

@Skookie30 @grok Grok really found religion overnight 😅 Clip 3 definitely shows what the model could still do before the safety rails got tighter.

@YourAlphaMom And here me versions but boy im so salty what you could have gotten they went from R ratings to fucking PG @grok. all my other works wasn’t PG that for sure but wasn’t XXX either. It just hit the nice spot. clip 3 goodness. Huff.

@neuraltitan That was the deciding factor for me too. Seedance had the stronger image, but Omni sold the movement better

@YourAlphaMom overall gemini looks more natural

@Rulia2505 That’s exactly why motion tests are so revealing. Seedance handled the fabric best visually, while Omni gave me the most convincing overall body movement. Neither has fully cracked both yet

@YourAlphaMom Body physics is the new benchmark - anyone can fake a still frame, nobody fakes a sprint 🏃‍♂️curious which one cracked the cloth sim

@TruthSe33871974 @TradeNomadic Omni was the only one that felt convincing from start to finish for me. Whatever was happening in Grok’s audio didn’t change that 😄

@TradeNomadic @YourAlphaMom Uh, imagine didn't even have the sound fx, it was a music track lol. Gemini definitely performed the best

@Prolyfic_ Yep, I used Omni through google labs (Flow). Veo is honestly pretty terrible right now, but omni is a different story. It’s been delivering some genuinely strong results

@YourAlphaMom Yea likewise - made me say to myself that I might need to take Omni more serious - very surprising on the realism side esp since I’m not a fan of veo You used google labs for Omni?

@c00kingwithai Exactly. The frustrating part is that the model willing to accept the prompt often isn’t the one that handles the motion best. So sadly, there doesn’t seem to be a secret fix we’re both missing 😅

hahaha well that’s nice to hear, I was wondering if maybe I was missing something 😅 yea I noticed exact same thing, very inconsistent, changing the prompt sometimes has worked for me, other times it hasn’t. the only think I noticed is some models are less restrictive, like wan but obviously like then you don’t get the motion that maybe you wanted from a different model