For the price this makes a lot of sense.
I'm not going to lie and say this is the most impressive deep fake program I've ever seen. I'd give it 4 Tacos, except for the price... It's a very good price point. $59 for 20 minutes of generated video a month? That's totally worth it.
So now that I have the program I'll answer my own questions I posted earlier:
1) Can you upload your own voice-over rather than use text to speech? YES! You can upload your own voiceovers and have the actors auto-synch to them. It works and doesn't sound robotic.
2) Can you upload your own footage and generate an actor? YES! You can. I have over 80 hours of footage of my now wife from when we were younger and she played a barbarian queen in our college feature film we worked on. About 30 hours of it was on a green/blue screen set. So this should be fairly easy to make her a surprise.
What would I like to see added?
There are programs/systems that will let you send in hours of voice samples and will compile a text to voice that sounds like the person you fed it. If they could improve the inflection of their text-speech voices to sound less robotic and add this quasi-common (but costly) feature, then man would this be a killer system!
Ankush_DeepWord
May 9, 2024Thanks so much for the review Gelobryn!
Yes, we are definitely working on voice cloning for users. The current issue is that it requires around 15-20 hours of high fidelity segmented audio to create a realistic voice model for use with TTS. However, we are working to lower this ultimately to around 20 minutes. However, if anyone every comes out with an API that meets our standards for a realistic sounding version of the voice cloning you described, we'd be more than happy to integrate with them so that our users can begin using such a feature sooner.
Thanks again,
Ankush
DeepWord