The thing the urks me about this is not so much the ML, but... who actually wants personalized video ads? I think it is very uncomfortable for people you don't know to say your name like they know a lot about you. This would be be awful to me even it were paid actors, costs be damned. This gives the privacy problem a human face and, imo, makes it much more discomforting. What kind of marketing context makes it feel appropriate to send a personalized video message that will be warmly received but also cost prohibitive to actually make one?

Not familiar with Loom or Vidyard, but really this whole premise seems really creepy, neural nets or otherwise. It smells like the the ads that proclaim HOT SINGLES WANT TO CHAT TO BOYS IN {your age range} in {your town}.

It was completely unnerving and creepy af.

The fact that this startup wants to automate this creepy invasive behavior tells me that the founders need to do some serious soul-searching about why they want to create the world they're working to create.

The video doesn't have to be personalized to your name or contextual information about you specifically, although that is the most exciting prospect for most people. It could be a topic you're interested in being visualized.

We're working with an e-learning company for example to help instructors create their courses more easily in their own voice. To me that's a more exciting use case. If you're interested to learn more about how our roadmap is focusing on use cases such as this feel free to reach out.

Right in your description you say:

> [People] want to see you, see that you are real, they want to hear your voice.

Yet that's exactly what this product takes away.

So the only way for this company to succeed is if their videos are so good that the targets never realize they're fake. I'm no entrepreneur, but it doesn't seem ideal if your business model is literally based on lies.

Not to mention, if the technology is that good, then it will be trivial to abuse this for corporate espionage, phishing, blackmail, slander, etc. Sure, if they don't do it, someone else will, but that's never been a valid excuse.

To us this is the next evolution to something like MailChimp or as another commenter mentioned- handwriting style fonts in marketing material.

As for abuse- we are heavily focused on making sure our platform is not abused. There are multiple checks in place to make sure only you can generate videos with your voice/face.

And I'm sure you've thought about the abuse potential and you're doing what you can, but what's going to happen when accounts get hacked? Companies already have to train their employees to be wary of emails impersonating their CEO ("please wire me $10,000 immediately for a high profile account"), and it's going to be much harder to train everyone to mistrust video or even live video calls if Tavus works quickly. I guess it's inevitable, but again, you don't have to be the ones to bring that future to life.

For example we’re working with a telehealth company to provide weekly/monthly health updates that they otherwise would send through emails and people would ignore.

I’m personally more interested in these product use cases than sales/marketing. This is also why we require set templates that limit what you can change.

We are also working on the flip side of this problem. In addition to building this platform Tavus is also building detection tools as we continue to explore what disclosure should be required on these videos. I know in that sense we’re both fire fighters and seem like we’re arsonist. However we’re really focused on providing a tool that builds immersive experiences while also fighting bad actors from other platforms that precede us/will come after us.

Maybe it would help if you included testimonials from people consuming the videos, rather than just producers. At the moment I genuinely can't imagine anyone feeling anything but misled and disappointed upon discovering their "personalized" message was actually auto-generated, but if I'm wrong, I'm wrong. If you have feedback from viewers of your videos (not just sales people seeing demos, actual end recipients) that understand how they were made and still like them, I think that would help demonstrate some positive value.

Edit: Maybe language learning? I wonder if hearing myself speaking a new language would help me learn it.

Language learning is a super cool use case! We're working on other languages atm

Sales unfortunately does have unsavory aspects to it that we at HN especially dislike

What is less clear from your website, is are you just handling the generation of the video, or their distribution also? Is there any integration with ad platform? Could I just send you an excel spreadsheet with video variable and a target email in the column?

will be interesting to see who writes them a check :)

The background content is customizable and dynamic per video, you can pass in a different URL for each video.

Currently we are only handling the generation of the video- not distribution just yet.

No, its embarrassing and would surely put the fear into anyone remotely technical whose org is considering adopting.

I'm curious how this ultimately scales. I'd bite on believing that these videos can convert better than a boring templated email, but I'd also assume most of that lift is because they're different and enough users haven't caught on to the fact that these are generated.

When people talk about personalized email working better than automated outreach, they usually mean actually personalized emails, which contain at least one sentence that is not just placeholder substitution. Templated emails with substituted variables are usually easy to spot, as evidenced by the number of "I'm really interested in what you're doing at Stealth Startup" messages in my email/LinkedIn.

So, to my question: do you think the lift in conversion you've seen in early tests is sustainable? As people start to understand that this is the new marketing tactic, do you think the average person becomes more aware of it and starts to believe that most videos in their inbox are probably just auto-generated also? Ditto for other use cases. If people like their video-generated product walkthrough because it feels personal, what happens when they start to realize it's almost certainly not?

It's a great novel; but this idea is pretty horrifying.

I personally would feel it is less manipulative and think of it higher.. but perhaps your goal is so people never find out?

Do we really need more manipulation in the world?

Would you rather be told the content is generated before or after you see the video?

There are also use cases that were working on with tele-health companies, allowing providers to send updates that build trust with patients that they otherwise wouldn’t have time for given the scale.

There are a lot of use cases to build more immersive experiences!

There are also so many cool use cases beyond sales!!

So perhaps that's my question -- how are you thinking about anti-fraud and anti-abuse at a high level? There's no HIPAA or CCPA for deepfakes yet, but one imagines that's only a matter of time. How are you planning to address this? Do you have any interest in having a hand working with lawmakers to develop this legislation to ensure responsible usage?

Currently our platform has a number of mechanisms to prevent abuse. We require explicit voice disclosure to perform any training on a voice/avatar as well as ensure ownership of a voice/avatar and resulting content is in the hands of the individual, not a company. This is in addition to limits put on how you can utilize our platform and what templates you can make.

We totally expect there to be regulation for technology like this and we'd love to work with legislators to enforce responsible usage.

We're working on the flip side of this technology as well- we're developing tools to detect generated video content in order to promote safer usage in the future.

How about if you wanted to teach a subject and your students wanted to see your face. Perhaps you could use this to just upload a bunch written lessons and this would help everyone.

Or how about using this for tv news/documentaries.

I cannot speak for Tavus, but this is awesome tech if used properly.

Great Job!

Whenever it starts with "Hey Josh So I found out that the playgound does not seem to limit what you can type in the boxes", you know it'll be a good one :)

Starting a new game of the Witcher? Instead of personalized text the cut scene is Geralt talking to you directly :)

With sales though it can be hard to see how the chicken nuggets are made.

The massive opportunity here is the conversion of existing low engagement mediums to high engagement voice and video which would include the following verticals (education, content creation, feedback, to name a few).

I understand where the negative comments within the sales / marketing use case come from, but it's important to realize the main KPI in that vertical is conversion, and sales orgs could care less about any other metrics aside from that. This may come at the cost of consumers treatment, but this happens within sales already with predatory practices such as hypertargeting, etc.

And obviously any technology within the umbrella of deepfake can be misused, so I'm less concerned about those types of comments.

How does this work? Like, it's literally just the same spam, but with a video. Do people really find this to be more useful ? What makes someone click on a video when they wouldn't have read your email, or clicked a link? To me, it's all the same low quality cold outreach.

Sales is a key part of every single company. Sometimes it can be seem unsavory when we see the reality of how it works.

However, I think the technology could be great for personal assistant apps, reminders/scheduling software, news feeds, and games/entertainment (especially). Anywhere a talking head could fit in a UI... as long as it's clear it's artificially generated. That is key.

As well- we'll dynamically generate background content too!

However we are approving every use case manually to make sure it’s not crazy.