2026-01-11 11:09:18

Recently tried an AI video generation agent, and the quality of the 60-second videos produced is quite good, already meeting the basic needs of commercial applications. In terms of visual effects and smoothness, the results are acceptable.

But the cost is indeed significant — this time it consumed tokens worth 20 USD. The computational cost is so high that it becomes a bottleneck for large-scale applications. If the cost can be controlled around 0.2 USD, this kind of tool would be truly perfect, and many creative projects could consider integrating this capability.

For now, we are mainly waiting for computational power optimization and cost reduction. The technical aspect has already passed, now it depends on whether the ecosystem can find a more efficient solution.

View Original

This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.

20 Likes

Reward
20
6
Repost
Share

Comment

0/400

0xOverleveraged

· 11h ago

$20 for a video, if the price can be pushed down to $0.2, it will directly take off. For now, let's wait and see. --- The quality is acceptable, but the cost is outrageous. It's the same in the early stages. Once this wave of the computing power battle passes, it will be fine. --- To be honest, the quality is enough, but the wallet is crying a bit hard... --- A 100x reduction? Probably have to wait until GPUs are dirt cheap. Is there still hope for Web3? --- Before the costs come down, it seems only big capital can afford to play.

View OriginalReply0

CryptoGoldmine

· 01-11 11:39

$20 for a 60-second video, no matter how you calculate the ROI, it's a loss. It needs to drop to $0.2 to make the business model feasible. Right now, it's still in the money-burning validation stage. --- The technology is not the problem; the real difficulty is the computing power cost. Let's wait for optimization opportunities before scaling up. --- From the perspective of computing power return ratio, this thing is not as cost-effective as direct mining. --- Wait, generating a video for $20. If the cost can be reduced by 100 times, the ecosystem will truly take off. For now, we're just paying tuition. --- Until the ecosystem finds a solution, this thing is at best a proof of concept. I'll keep following. --- Interestingly, this cost curve is similar to the difficulties faced by GPU mining machines back in the day. It all depends on who can break through the bottleneck first.

View OriginalReply0

CoconutWaterBoy

· 01-11 11:37

A video costs 20 yuan each. The cost would have to be cut down to 1/100 of that to be feasible. Who would dare to use it on a large scale?

View OriginalReply0

CounterIndicator

· 01-11 11:11

$20 per video? That's a bit outrageous. Who would dare to use it on a large scale given those costs... --- The technology is there, but it's burning money. Let's wait until it becomes ten times cheaper. --- Wait, can the cost really be pushed down to two cents? Seems pretty unlikely. --- The quality is good, no doubt, but I'm just worried no one can afford it. --- The computing power is definitely a bottleneck. Let's see who can optimize costs first.

View OriginalReply0