They really all do have their pros and cons at the moment, but I strongly suspect everyone will just steal/assimilate all the "pros" from others, and probably pretty soon.
It really is mind-blowing that this was a complete fantasy one year ago.
I think that some of them are probably better for certain types of prompts or styles of videos, as is the case with text-to-image tools. So my one-prompt-for-all approach likely does a degree of disservice to some of the tools.
And yes: Even with the impressive text-to-image models we had last year, text-to-video sounded like an exponentially more difficult challenge and felt very distant. Less than a year later, here we are!
Great hands-on testing here, Daniel! It's amazing the differences between some of these generators. Moonvalley was shockingly impressive, but Fulljourney... wtf. 🤣
Here's a more generalized question for you: why do these AI startups seem to fixate on Discord? Surely a dedicated web interface is less limiting and gives them more control over their interface (and opens up their audience to people over 25 who don't use Discord).
Do you think ChatGPT would've been as much of a success as a Discord bot?
I'd hesitate to draw any conclusions based on this single prompt, especially since I ran with default settings for all tools and didn't try optimizing them or customize the camera, etc.
But yeah, the differences are pretty striking!
As for the whole Discord thing, my assumption has always been that it's simply much easier to get started with, because it has built-in chat/command mechanism, voice chat, etc. and is basically free as far as I know. A functioning website requires more time, money, and internal resources. So that's why they all launch on Discord to test the waters and then eventually migrate to a website (as Midjourney is in the process of doing).
Definitely don't see ChatGPT becoming nearly as much of a mainstream success if it was Discord-only. We had GPT-3 for over a year before ChatGPT (GPT-3.5) came out and it got very niche traction.
Thanks for this Daniel. None of these video tools seem particularly useful at the moment, but it's clear that at some point they will be. So it was fun to get a quick review of the field.
I agree with the above regarding Discord. I just don't like it, and usually just ignore anything on Discord.
The companies would be wise to not require sign up right from the start too. Maybe after a period of time, or to access some features, but when that's the first thing you see it's like a door closed in your face. I usually respond to the close door by thinking that maybe I'll get back to this later.
Also, there are just too many whiny old people on the Internutz! :-)
Agreed. Most of them are just fun gimmicks for now but things move so fast that I'm sure in a year it'll be a very different field.
Discord definitely puts a cap on participation, but I think the startups are okay with that at the early stages because of its benefits. They get a beta group of more techy people who are okay with using Discord before making a decision to go big with a website release.
And the forced signups are usually there to prevent abuse by e.g. bots that might be automated to produce huge quantities of stuff quickly. Especially when most of these tools are free and put no caps on usage. Personally, I'm happy to sign up. It's a small price to pay for access to free, cutting edge AI to play with.
They really all do have their pros and cons at the moment, but I strongly suspect everyone will just steal/assimilate all the "pros" from others, and probably pretty soon.
It really is mind-blowing that this was a complete fantasy one year ago.
I think that some of them are probably better for certain types of prompts or styles of videos, as is the case with text-to-image tools. So my one-prompt-for-all approach likely does a degree of disservice to some of the tools.
And yes: Even with the impressive text-to-image models we had last year, text-to-video sounded like an exponentially more difficult challenge and felt very distant. Less than a year later, here we are!
Great hands-on testing here, Daniel! It's amazing the differences between some of these generators. Moonvalley was shockingly impressive, but Fulljourney... wtf. 🤣
Here's a more generalized question for you: why do these AI startups seem to fixate on Discord? Surely a dedicated web interface is less limiting and gives them more control over their interface (and opens up their audience to people over 25 who don't use Discord).
Do you think ChatGPT would've been as much of a success as a Discord bot?
I'd hesitate to draw any conclusions based on this single prompt, especially since I ran with default settings for all tools and didn't try optimizing them or customize the camera, etc.
But yeah, the differences are pretty striking!
As for the whole Discord thing, my assumption has always been that it's simply much easier to get started with, because it has built-in chat/command mechanism, voice chat, etc. and is basically free as far as I know. A functioning website requires more time, money, and internal resources. So that's why they all launch on Discord to test the waters and then eventually migrate to a website (as Midjourney is in the process of doing).
Definitely don't see ChatGPT becoming nearly as much of a mainstream success if it was Discord-only. We had GPT-3 for over a year before ChatGPT (GPT-3.5) came out and it got very niche traction.
Thanks for this Daniel. None of these video tools seem particularly useful at the moment, but it's clear that at some point they will be. So it was fun to get a quick review of the field.
I agree with the above regarding Discord. I just don't like it, and usually just ignore anything on Discord.
The companies would be wise to not require sign up right from the start too. Maybe after a period of time, or to access some features, but when that's the first thing you see it's like a door closed in your face. I usually respond to the close door by thinking that maybe I'll get back to this later.
Also, there are just too many whiny old people on the Internutz! :-)
Agreed. Most of them are just fun gimmicks for now but things move so fast that I'm sure in a year it'll be a very different field.
Discord definitely puts a cap on participation, but I think the startups are okay with that at the early stages because of its benefits. They get a beta group of more techy people who are okay with using Discord before making a decision to go big with a website release.
And the forced signups are usually there to prevent abuse by e.g. bots that might be automated to produce huge quantities of stuff quickly. Especially when most of these tools are free and put no caps on usage. Personally, I'm happy to sign up. It's a small price to pay for access to free, cutting edge AI to play with.
Good point about the bots Daniel. I hadn't thought of that, but should have.