Have you ever wondered how much unseen footage gets lost forever since no one cares for a take that does not end up in a final Movie or Commercial. Well I used AI to imagine some of the takes that may have happened and we will never know.
The original Commercial was a flash-based interactive page that featured our main character and moderator presenting a floor mat for cars with the distinct feature of being easily cleaned whenever it gets dirty. The interactive part put you in the drivers seat and you could choose which mess was dumped on the mat and consequently on the beautiful lady presenting it.
It is the first time I used Veo3 for generating parts of the video. But to preempt a lot of the questions. My workflow got even more complex to achieve this outcome. A lot of tools and a lot of Know-How go into generating a video like this.
Excellent. I think really show cases what the best current workflows can achieve. VE3 is not cheap to use to unlock it's full potential but the tool is developing well. I've been playing around with VE3 via Flow but not having last frame and ingredients available for VE3 limits it to an extent. Will be interesting to see how much more consistent it gets when they release. I'm presuming you are manually creating 1st frames from the previous video last frame and then stitching the videos together externally. Hopefully that won't be needed in time.
I think WAN2.1 still edges VEO3 in some respects with a well trained LORA for messy but the overall quality of VEO3 is moving way ahead of open models. The integration of speech I don't think is going to happen soon.
As always, the lack of control over safety guardrails is frustrating with private models. Google is more lax than OAI but I still run up against silly barriers even when generating safe videos. This may not impact some users but it does limit creativity even if NSFW is not the main objective.
Definitely and exciting time. I suspect VEO4 will be available by the end of year and VEO3 tools will be improved.
Really impressive. Yes, I can tell it's AI, but the still decent realism, coherence, and audio matching was very satisfying. I know a heated debate about AI is whether it actually constitutes creativity, but it seems like you've had to do a substantial amount of planning and input to create this from an initial bright idea.
Love you, too
