Comparing Wan, VEO, SORA, LUMA and GROK with a paper boat navigating on a mountain torrent.

Prompt used for all models:

Make a video of a paper boat descending and whirling in a natural torrent water stream from the Alps.

A few natural rocks are in the torrent, the torrent is small and the banks are made of herb, flowers, small bushes, in spring season.

No wild life.

Just the boat descending on the stream, with realistic physics for lights, water and boat movements: it whirls, slows down and goes faster depending on the current.

⏩ all generations presented below are first try

⏩ Except for LUMA, geenrated videos come with sound. So please swith the sound ON on the videos.

RAY3 (by Luma Labs AI)

• Time to generate : about 1 minute if I remember correctly.
• Where? on a free plan on Luma Labs’s dream machine: https://dream-machine.lumalabs.ai/
• Evaluation: my paper boat is sailing upstream! 🤦‍♂️ Also, no sound.
• Remarks: I made a second try with a prompt asking the boat to sail more naturally, downstream.

WAN 2.5 (by Alibaba)

• Time to generate: at least 20 minutes . It took so long that I closed the tab and came back several hours later
• Where? on a free plan with https://create.wan.video/generate
• Evaluation: great fidelity to the prompt, level of details, great sound. The movement of the boat is not really coherent though (should not be so idle given the current)

GROK (by xAI)

• Time to generate: 10 to 20 seconds.
• Where? on a free plan with https://grok.com/imagine
• Evaluation: no photo realistic, movement of the boat is not natural, resolution is low, sound is compressed

SORA 2 (by OpenAI)

• Time to generate: between 5 and 10 minutes if I recall correctly
• Where? on a free plan with https://vidgo.ai/
• Evaluation: photo realistic, camera angles feel elaborated, movement of the boat is so natural, resolution is high and even the blur feels as if an amateur would have shooted out of focus. Sound is realistic and detailed though not perfect.

VEO 3 (by Google)

• Time to generate: a few minutes if I recall correctly
• Where? on a Google pro plan with https://labs.google/fx/tools
• Evaluation: photo realistic, movement of the boat is so natural, resolution is high. Sound is realistic, really good.

Conclusion

Rankings

• There is a tie between SORA 2 and VEO 3 in my honest opinion.
• WAN 2.5 is just behind, not because of the quality of the video but because the boat movements are not really natural.
• GROK and RAY3’s quality is frankly below in terms of photorealism, level of details and physics.

The case of GROK

GROK’s approach is distinctive: first, it generates in seconds literally dozens of images following the prompt, all with slight variations. The user is then invited to choose one of these frames as the preferred starting point to generate the video. The video is created at extreme speed (10 to 20 seconds or less in my experience). The video is produced in vertical mode and has a discreet but visible anime style, making it popular for shorts posted on social media.

Other models, newer models?

This comparison was made in relation to a list I maintain on AI apps for visual creation: image, video, dubbing, 3D models and more. This list is made to save you time by providing a frequently updated, one-stop landscape of AI apps for visual creation, grouped in meaningful categories.

About Me

I’m an academic and independent web app developer. I created nocode functions 🔎, a free, point-and-click tool for exploring texts and networks. It’s fully open source. Try it out and let me know what you think. I’d love your feedback!

• Email: analysis@exploreyourdata.com 📧
• Bluesky: @seinecle 📱
• Blog: Read more articles 👓 on app development and data exploration.