Inspiration
Audio deepfakes have achieved an unprecedented level of realism. Today, commercial tools to create audio deepfakes are now available for as little as $1 per month, enabling the generation of highly convincing and lifelike audio replicas. Scammers, for example, have been exploiting audio deepfake technologies all around the world (including Europe and Latvia) to increase the level of sophistication and believability of scam calls, misleading people with fake voices of their loved ones in distress for extortion or financial fraud schemes. Nowadays, entry level to voice cloning became so low, that even common people can do that themselves.
What it does
Perfect Paul is an AI based anti-spoofing tool for the common common people. It detects the AI generated voice and notifies about it the user whilst being easy to use, accessible and affordable.
How we built it
We recorded the voice from one of the organizers of hackathon. Trained AI Model on voice synthesis, and the detection of So-called artifacts that are being produced from voice clones. We are able to attack these artifacts. And are able to also to test on multiple different other levels. Like background noise. And, The original uploaded test voice. So it's an app-based solution which only needs an API connection to test against. Which means all the voice samples are safe and in your hands. A data breach wouldn't allow for for hackers to steal. All the voices of the user base.
Challenges we ran into
-lack of problem specific mentors -unable to get access to Microsoft azure (US) -inability to get the A.I. clone fully functioning during live call with cloned voice A.I.
Accomplishments that we're proud of
-forming a strong, multitalented team -making an AI voice clone from a real person
- using ai as a wrapper to establish a connection between Cloned voice Model That is using GPT Turbo To interact through a twilio voiceover IP connection through the uk for number.
What we learned
-Much like Tony Stark in Iron Man 1, how to build something with a bunch of scraps while being stranded in the desert. At least that's how it felt like after being denied access to multiple tools due to various reasons.
What's next for Perfect Paul
-training our own AI voice detecting model -building MVP
- getting access to Microsoft Azure (US) A.I. API
Built With
- davinciresolve
- elevenlabs
- figma
- openai
- twilio

Log in or sign up for Devpost to join the conversation.