Google Veo 3

Let's have a look at some amazing sample clips made by others with Veo 3...

Many more examples, continuously updated, can be viewed on Google Flow TV, with the advantage that in every case it is possible to see the prompt that generated the video (but don't forget that the same prompt will never give the same video due to the random nature of AI).

As of July 2025 Google Veo 3 (deepmind.google/models/veo) is considered to be by the best AI video generator for short photo-realistic video clips in several ways:

• its output is usually very life-like, if that is the desired result (of course it can also do animations etc)

• it can add sound effects and voices in a single pass, even singing

• it can usually make text appear in the video correctly, whereas other AI video generators usually display gibberish

• it displays fewer glitches (morphing etc) than most other AI video generators

• it understands the laws of physics better than most models, so motions appear relatively natural

• the processing of the video happens fairly quickly, often in about a minute, whereas some other high quality generators take much longer. In my experience the 'image to video' mode can take significantly longer, however.

• although the output is somewhat random (the same prompt will not generate the same video) , Veo 3 is considered to be more consistent than other AI models, so the same prompt will generate a relatively similar video.

• it is possible to have a seed image that will become the first frame of the video (it's still necessary to give a text prompt to control the motion of the characters and the camera). This meant we have more control over how the video looks and makes it very easier to have consistency in the characters and environment from one video to the next. 

Its main disadvantages as of July 2025 are that

• it can be more expensive than other AI video generators, but it doesn't have to be. The best way to use Veo 3 is to get a Google AI Pro subscription, as I have done, which costs about €20 per month. For that you get 1 thousand credits and it costs 20 credits to make an 8 second Veo 3 video in the lower quality (but still excellent) ‘fast’ mode or 100 credits to make a video in ‘quality’ mode so you can make 50 videos per month in fast mode or 10 Veo 3 ‘Quality’ videos. So it costs about €0.40 to make a video in Veo 3 fast mode, the mode I have always used (the difference in quality isn't huge). You have to remember though that quite a few of the videos you make will be unusable for one reason or another. Avoid the Google AI Ultra plan that costs $250 per month!

• it sometimes displays burnt-in subtitles that were not asked for and which are incorrect. If you find that to be a problem you can include sentences like this in the prompt: "No subtitles, no captions, no written elements or text should appear on the screen at any point. The screen must remain clean — only visuals and spoken dialogue." It is apparently also helpful to avid putting quotes around any text that is to be spoken within the video. Use a colon instead. Some sites have an option of using AI to remove burnt-in subtitles.

• the videos it makes are always 8 seconds long. Veo 2 allows videos to be extended but does not do sound. I expect Veo 3 will get the extend option some time soon.

• it does not have some of the features that other AI models offer, such as more control over the length of the clip, the aspect ratio, the resolution, the possibility of including 'elements' that are to be included in the video etc

As of June 2025 I understand there are three ways of accessing Veo 3:

• via a Google AI Ultra plan. This option may not currently be available in the EU but is available in about 70 countries including the US and the UK. I believe it costs $250 per month! (Half price for the first three months.)

• via a Google AI Pro plan (free for the first month, then €22 per month as of July 2025). Check out this good video from Kevin Stratvert for instructions. He will point you HERE. This is the option I recommend.

• from within a package such as Pollo.ai. This is how I made my first AI video, the Buxton protest. Other packages that include AI are Freepik AI and Deevid. I'm expecting it to soon also be on InVideo soon. 

• From a Google Cloud account. This is a very attractive option - you get $300 of credit, valid for 90 days, and I suppose you can then switch to a different Google account to get the same credit again. Here is the help I got from Gemini with that. This is how I made the 'Wardlings in space' clip below (June 2025). For the full story behind this clip, click HERE. However, this option currently does not allow you to do 'image to video' which I find to be the most interesting option, so I barely use my Google Cloud account any more and do all my work within the Google AI Pro plan.

My main aim with AI video generation was to be able to make a longer video, 3-5 minutes in length, with  a proper plot, probably on the theme of a quest. The first clip for that story is below and you can follow the development of other clips HERE. 

I've had some trouble trying to get friends and family interested in AI video making to the extent that they would get involved themselves. At the time of writing my only success has been to get a proposition of a prompt from my brother Peter who suggested this one word prompt: "Holism" which for him is the idea that the whole is different to the sum of its parts. But what on earth would Veo 3 make of such a brief and abstract prompt? Here is Veo's take:

It's very difficult for Veo to generate a video from a single word, and an abstract word at that. I'm not convinced that its output is a good response to the word. Also, I feel that the arm doesn't look very real, though the forest is nice enough. Finally, I always feel rather cheated when Veo generates video with black bands top and bottom, since it steals away some of the resolution. 

A couple of days ago (in July 2025) the Lionesses, the UK's female national football team, won the Euros. It's the first time a UK national football team has won a major tournament on foreign soil, so they deserve recognition by me in the form of me hugging a lioness. This video is made from a photo of me in the Valle di Genova in the Italian Dolomites. I'm very happy with it.

The reality is that there aren't any lions in the Dolomites, but there are bears! So I also had to make a video of me being chased by a bear (it's a grizzly, but really of course the bears in the Dolomites are the smaller black or brown bears).

With all that foliage (how many leaves do you see?) it's a complex scene to render, and Veo doesn't have enough computing power to render the bear's rough fur very well on top of that. When the camera moves to follow the action that places even more demands on Veo, and the quality of the video at the end is not great. Generally, the more dynamic the video, the lower the quality of the rendering will be. Let's not forget that so far I have only ever used Veo 3's 'Fast' (cheaper, lower quality) mode, rather than the 'Quality' mode that costs five times as many credits per generation. Worst of all, the video makes me look even fatter than I really am. Also, why did Veo add a second bear? Still, not bad, I think. Here's the prompt I used:

A grizzly bear enters the frame on the path behinds the man. It stands up on its hind legs and roars. The man turns around, sees the bear and then runs away to the left, chased by the bear, while the camera tracks them both. There is light wind. We hear the sounds of the forest including birds and the babbling stream.

I invite you to now discover my 'major project' to create a three minute video called The Quest. How can the making of a three minute video be a 'major project'? Well, there are so many variables that you can potentially control in Veo 3, and so much trial and error involved, especially in trying to make the characters look consistent from scene to scene, that this project will probably take three weeks to complete!