Audio Experts Deconstruct Deepfake - Can You Tell If a Voice is Real?

How can you tell if audio is real or deepfake? AI voice cloning tech is good but can we tell it apart from real voice? Audio experts Mike and Izabela Russell investigate a controversial viral clip of UK opposition leader Keir Starmer to determine if it is real or a deepfake using AI voice cloning tools and audio restoration techniques.

Key Video Moments:

0:00 Intro
0:48 Original Viral Clip
1:13 Discussion on Deepfake Voice Clones
6:35 Attempting to Recreate Deepfake Audio with AI Tools
10:33 Editing Deepfake Audio
15:18 Using Audio Resroration Tools to Analyze Deepfake
19:35 Using Enhance Speech AI Tool (Shocking Result)
22:42 Using an AI Speech Detector Tool
23:39 Conclusions

----

🎙Cool Stuff For Audio Creators🎙

----

Hi there! I'm Mike Russell from Music Radio Creative. I am an Adobe Certified Professional and my aim is to teach you the secrets you need to know to create better-sounding audio with the use of various techniques. From Adobe Audition tips and tricks to your choice of microphones, audio gear, home recording setup, and many more! Podcasters, Livestreamers, Filmmakers, and even YouTubers can make use of what I am about to show!

----

#deepfake #ai #aivoice #aivoiceclone #aideepfake

[Paul_anderson_creative]

I have never seen such a detailed and thorough examination with such clarity. Excellent step by step work and explanations..Kudos 🙏🙏🙏🙏🙏

[Orhan-wcib]

how much harder would it be to make it like a real conversation. say a ai generated discussion between a political leader and some of their associates. I assume it would be a lot harder?

[AG_before]

My comments keep getting deleted Mike, Izabela. Are there YT rules and regulations I'm missing?
Well, great video then. Thank you for sharing this. It actually _really is_ a great video. I believe transparency will be the key. Thanks. 👍

[TheHermit]

Not to be the harbinger of doom, but the fear is valid. These tools have the capability to serve as a powerful tool for propaganda and then some.
The deeper this goes, the more everything we see, hear, and read becomes illusory.

A few months ago, I used ElevenLabs for a video game project I've been working on (to save money on hiring voice actors). The audio I generated received light editing, splicing, and general clean up. After watching your video, I decided to run some of the audio through the AI Speech Classifier - I was actually unaware of its existence. Interestingly, only portion of the output was identified as 97% AI generated, while a significant number fell in the 0-2% AI generated range i.e. "unlikely to be AI". Again, this classification was made after making only light edits using RX9: general audio clean up without any background noise. It's a brave new world. 🙄

[ryanhare]

Great video. For me the emotion in the voice is the give away between neural networks generated and not. We are in dangerous times with how powerful and accessible these tools are and how much doubt they can sow.

[instructionaldesign]

Always enjoy your insights and always learn a lot, so thanks for this. The question is, "Is this video a deepfake?" :)

[AshisMahato-osxl]

Sir i edit a voice in audacity and the voice is sounding very good. But now i want to make another voice like previous voice but i can't, and i try to make a preset of previous voice editing. So how can i make a preset in audacity from previous save voice editing.. please help

[QuirkyCactus-nqmf]

I’d like to know if the video of Jason Momoa in China speaking Mandarin is real. I don’t think it is. It looks and sounds odd to me.