AI Video Marketing: Is this the future for dealership sales and service?

[slatqh]

Sounds cool. I did be interested in seeing some demos, see what you manage to create! Could you publish a few demo videos like I did and try to top them? Lets see how far we can take Veo3.1 and collaborate.

Cool. For now, I’m just exploring options and checking if there’s real demand for this kind of service. I’m not a car enthusiast and could be wrong, but it seems like VEO3 struggles to generate realistic car videos, especially the dashboard, which I think isn’t acceptable for advertising. This is a technical challenge I want to explore further and see if it can be solved with the current state of technology.
btw, is there any standard for video length? VEO3 made an 8 sec clip, which feels too short. I’d say the sweet spot should be around 30–60 secs.

 

[Esteban]

Cool. For now, I’m just exploring options and checking if there’s real demand for this kind of service. I’m not a car enthusiast and could be wrong, but it seems like VEO3 struggles to generate realistic car videos, especially the dashboard, which I think isn’t acceptable for advertising. This is a technical challenge I want to explore further and see if it can be solved with the current state of technology.
btw, is there any standard for video length? VEO3 made an 8 sec clip, which feels too short. I’d say the sweet spot should be around 30–60 secs.

I did say do not work in something that you are not truly enthusiastic about, you will really want to love cars and trucks as much as everyone here to get car videos right. That said, you can use the same tech to make videos just about anything, but focus on something you love. Say bikes, or boats, or something you love. Veo3 does struggle , you need to use Veo3.1 pro, not the flash version. Not only dashboard remains challenging, also badges, text, other small things. There is a lot of other challenges you will find once you start writing the code, but I think all of them will be cleared within the next few months.

As for the length, yeah, around 30 seconds. If you make them too long, they dont work well with youtube shorts, tiktoks, and insta reels, the limit is 1 minute, that is asking for a lot of attention now-days. You want to engage the users in the first few seconds, and you will be lucky if you get a full minute of their attention. Clips can be configured to be between 5 to 8 seconds, and then you can extend them a few times. So if you wanted them to be a single shot of more than 8 seconds you can do it, but they dont usually work to well. I did say that you want to keep the clips short, and then mix angles, go into the interior(for cars), do detailed sections, etc... You can apply this to just about anything. Think real estate, vacation places, any product for sale...

Videos definitely provide an edge for sales, specially if your competitors are not doing them. There is a huge market, yet we need to figure out how to make it all look good first, so they convert.

 

[slatqh]

I do love cars and ride motorcycles, but what I mean is that I’m not a car dealer and don’t really know how the industry works from the inside. I’m just looking for opportunities to build a startup business. The reason I’m focusing on cars is, first, because it’s something I’m genuinely interested in, and second, even as foundation models improve (which they certainly will), they’re still not trained on specific car interior datasets.


I could train a model myself or at least allow dealers to upload images temporarily and use those to fine-tune the model so it can produce the desired results. That’s where I see potential. But as I said, for now I’m just exploring and trying to figure out whether there’s actual demand or need for such a service.

 

[Esteban]

I do love cars and ride motorcycles, but what I mean is that I’m not a car dealer and don’t really know how the industry works from the inside. I’m just looking for opportunities to build a startup business. The reason I’m focusing on cars is, first, because it’s something I’m genuinely interested in, and second, even as foundation models improve (which they certainly will), they’re still not trained on specific car interior datasets.


I could train a model myself or at least allow dealers to upload images temporarily and use those to fine-tune the model so it can produce the desired results. That’s where I see potential. But as I said, for now I’m just exploring and trying to figure out whether there’s actual demand or need for such a service.

I see what you mean now. The idea of fine tuning a model is great, you take users pictures to create a LORA, and then use it to generate the video. Worth giving a shot, Google cloud provides some free compute you could use for training the model.