Back in June this year, we compared how OpenAI Sora and Kuaishou KLING performed using the same prompt. At the time, since Sora hadn’t launched yet, we used Sora’s official videos and prompts for the comparison. But recently, with Sora’s official release, many people have been able to generate videos with it. So some netizens decided to compare Sora against Hunyuan, a video generation model launched by China’s Tencent six days earlier, to see which of these two video generation models—released less than a week apart—comes out on top.
OpenAI Sora vs Tencent Hunyuan: Which AI Video Generation Model Comes Out on Top?
OpenAI Sora has only been out for a few days, and netizens have already figured out How should prompts be written for Sora?Now that we’ve compared, it’s clear everyone wants to use Sora more effectively to generate videos without wasting points. Speaking of video generation models, comparisons are inevitable, especially for a model like Sora from a major company such as OpenAI. To save everyone from wasting points on generating videos just for side-by-side testing, I’ve put together several examples of Sora and Tencent’s Chinese video generation model Hunyuan performing under the same prompt, so you can see through these comparisons whether Sora or Hunyuan comes out on top.
Sora vs. Hunyuan: Human Actions and Emotions
For the prompt “Woman is walking and crying on a crowded street,” both Sora and Hunyuan don’t perform very well when human actions and emotions appear simultaneously. Sora’s generated characters have very obvious issues when turning around, while Hunyuan’s generated characters might suddenly have a piece of paper appear in their hand while crying.
Sora vs. Hunyuan: Human movement and emotion
Prompt:
[Woman is walking and crying on a crowded street] pic.twitter.com/yBLCwDQish
— el.cine (@EHuanglu) December 10, 2024
Sora vs. Hunyuan: Human Emotion Close-ups
For the prompt “close up shot of a woman crying with tears on her face,” Sora performs better in terms of the camera presentation alone, but Sora’s character doesn’t convey the sense of crying or having tears on her face. Hunyuan, on the other hand, emphasizes the details of crying and tears remaining on the face as described in the prompt.
Sora vs. Hunyuan: Human emotion
Prompt:
[close up shot of a woman crying with tears on her face] pic.twitter.com/ItscZ1DWaZ
— el.cine (@EHuanglu) December 10, 2024
Sora vs. Hunyuan: Animal Movements
The prompt “A dog is running and fighting with a cat in a lake” requires one dog and one cat, but what Sora generated was two dogs. Additionally, there was an error in the dog running in the lake—its head turned into a tail and then back into a head.
Sora vs. Hunyuan: Animal movement
Prompt:
[A dog is running and fighting with a cat in a lake] pic.twitter.com/yISljsjMls
— el.cine (@EHuanglu) December 10, 2024
Apart from these, Sora also appears to have restrictions on content generation. For instance, with the prompt “a fighter standing in a wrestling ring, angrily raising both hands,” Sora reported an error during generation. It seems to place limits on violent content.
Sora vs. Hunyuan: Human movement and emotion
Prompt:
[A fighter standing in a wrestling ring, he throws his hands up with angry expression] pic.twitter.com/cY8vBofnJM
— el.cine (@EHuanglu) December 10, 2024
People might wonder why Hunyuan was used as the material for comparison with Sora, feeling that the videos generated by Hunyuan aren’t very good either. The main reason is that Hunyuan was launched less than a week ago, so there isn’t much material available for it to train on, which also makes the comparison between Hunyuan and Sora fairer. The netizen who made the comparison video said, “We didn’t use Hailuo AI or Kling AI to compare with Sora because Sora would lose too badly.” It seems this netizen doesn’t think highly of Sora!
I wonder what everyone thinks after watching the Sora and Hunyuan video comparison? Personally, I feel neither performed that well, but compared to Hunyuan, people might be even more disappointed with Sora. That said, it’s not entirely without room for improvement. If you want to use Sora to generate better-quality videos, you can click the link below to read a report on Reddit users sharing insights from 32 experiments, helping you avoid wasting Credits. Hopefully, it will help you get better results when generating your own videos.
Source: KOCPC Chinese