I would like to preface this post by affirming that what I did here, type a few words into a textbox and minutes later received a short video clip of what I typed, is nothing short of amazing. And it’s only getting better as these AI giants keep pushing the envelope. It’s daily that new records are broken as AI companies push out new, faster, better, more intelligently capable AI systems. So while I am harsh in my expectations and am critical of what I got, I am nonetheless impressed and awed by it.
I’ve been testing out Sora since I got access to it. I’m not sure if being a ChatGPT Plus subscriber gives me more videos on Sora but so far I’ve been able to generate as many as I want (which isn’t much). [I checked and yes, Sora has me listed as a Plus member.] I haven’t been very pleased with what I’ve generated with Sora, especially the two videos I’m sharing in this post. I had 10 free credits on Veo so I tried the same prompt on Veo that I used on Sora to see how it compared. With 10 Veo 3 credits I was only able to generate one video.
First prompt:
A group of Starfleet officers while conducting water quality tests on a small creek in the forest, run into two zombies. They try to stun them with the phasers as they run away trying not to harm the zombies.
Sora First Try
I was not too happy with the video. I could handle the strange movement of the characters but there were no zombies. That was the main purpose of the clip! I tried re-prompting to get zombies and I also had to prompt it to add a more diverse landing party.
Sora Second Try with Re-Prompting
Sora second prompt:
Starship landing party on a forested planet near a river fighting off and running away from zombies without actually killing them.
The movement was still strange but at least the above clip had zombies! And it didn’t need much diversity because it for some reason put helmets on the landing party.
Veo First Try (ran out of the entire 10 credits I strarted with)
Too bad I could only generate a 5 second video because this first try was great! The movement looked more natural, more realistic, the landing party fired their phasers, and there was sound! The zombies need to be better but what a great start.
I checked to see how much it would cost to create more videos. Here is Veo’s pricing:
I’m almost tempted to pay the $25 to play with Veo some more!
While looking up Veo I saw something about Flow, so I Googled it and here’s what Gemini generated for me, from my Google Search, and I was satisfied with its answer, so I didn’t click any links!
Flow is a new AI filmmaking tool from Google that utilizes Veo 3 and other generative AI models to create videos. Veo 3, on the other hand, is the generative video model that powers Flow, capable of generating realistic and high-quality videos with synchronized audio.
Key Differences and Relationships:
- Flow: An interface and tool designed for creatives to use AI to make videos. It allows users to describe their vision using natural language, manage assets, and edit existing clips.
- Veo 3: The underlying AI model that generates the actual video content, including visuals and synchronized audio.
How they work together:
- Flow uses Veo 3 to generate video clips and scenes based on user prompts and instructions.
- Flow provides a user-friendly interface to control Veo 3, manage assets, and edit generated clips.
- Gemini and Imagen models also contribute to Flow’s capabilities, including text-to-image generation and natural language understanding.
In essence, Veo 3 is the engine, and Flow is the dashboard and toolset that allows users to interact with and control the engine to create AI-powered videos.
For example, you can use Flow to tell Veo 3 to generate a video of a character walking through a forest, and Veo 3 will then create the visual and audio elements to make that happen.
I checked out flow and could not create anything for free. They were quick to offer me free trial if I subscribed right then and there. Here’s Flow’s subscription offer:
So I accepted their offer and am trying Flow for free for now. I got 1,000 AI credits and started prompting Flow.
Updated Prompt so it wouldn’t have copyright material related to Star Trek:
A starship landing party in a forested area conducting water quality tests on a flowing creek using sensors. Out of the woods come shambling at them undead zombies. The crew of the starship point their phasers, set to stun, at the zombies firing as they run away in the other direction.
No sound. I can’t figure out how to do that. I spent 40 AI points prompting and trying to get a video clip that looked decent. Something Flow, running Veo 3, can’t seem to do is distinguish between the landing party crew and the zombies. It kept making the crew zombies. Ugh.
So I tried expanding a clip because creating scenes proved confusing and trying to figure it out I accidentally deleted some videos that I spent AI points on and I don’t know how to get it back! Undo, Control-Z, did not work.
Here’s the expanded video with follow up prompt: The landing party outruns the zombies leaving them behind.
Very weird with more crew members as zombies. And still no sound.
So I tried going back to the source, Veo. Tracking down the link proved complicated, more than it should have been. I was given the choice to use Veo 2 with Google Studio or with Gemini Flash 2.5. I tried both with the original prompt I used on Flow.
Veo 2 with Google Studio
Not much better than Flow. It started much better than anything I got from Flow but in the end one of the crew was a zombie, and very prominently so!
Veo 2 with Gemini Flash 2.5
I finally got sound! Not sure why. This one isn’t bad and comparable to the Veo I generated at the beginning of this post. I have a lot to learn and now I’m down to 930 AI credits or points. What a costly way to figure these video generators out!
