Amanda Smith is a freelance journalist and writer. She reports on culture, society, human interest and technology. Her stories hold a mirror to society, reflecting both its malaise and its beauty. Amanda's work has been published in National Geographic, The Guardian, Business Insider, Vice, News Corp, Singapore Airlines, Travel + Leisure, and Food & Wine. Amanda is an Australian living in the cultural center of gravity that is New York City.
I've never been good at public speaking -- I share this fear with many. It all started in high school, when I had to present to the class but couldn't keep a straight face and got in trouble. The teacher misread my nerves for mischief, and I got penalized.
A couple of additional formative moments like this and I told myself that I'm not a good public speaker. The monkey mind is a good persuader, and I never challenged whether this was true. One-on-one, I'm fine. But my body goes into fight or flight at the thought of "presenting."
The good thing about artificial intelligence is that it will give me feedback straight. Yes, it can be a people-pleaser, spitting out responses in seconds and hallucinating. If I ask it to be brutal, it's probably better than a kind coworker. Friends don't want to offend.
I decided to use ChatGPT, just because I know my way around the AI tool well, but any chatbot would work. There are different ways to use AI as a public speaking coach. You can upload a video or a voice recording or talk to the tool about a specific area you need feedback on.
Can ChatGPT help me love public speaking? It's a tall task, but one worthwhile testing out.
I logged into ChatGPT and downloaded the mobile app for faster voice notes. Then I looked through some past call recordings (from story interviews or client calls) to see what I'd be comfortable uploading into AI. I avoided any sensitive information or confidential conversations. I would direct ChatGPT to focus on my voice, rather than the conversation.
To test ChatGPT (and my limiting beliefs), I shared a recording of a recent interview with a source and assigned AI a role/person to emulate -- Vinh Giang, the famous public speaking coach (and fellow Aussie).
My first prompt: "This is a file of a call I had with a source for a story. I want to improve my public speaking skills and confidence. Act as a public speaking coach, emulating communications coach Vinh Giang. Give me practical advice and help me see my weaknesses. Only focus on the female voice in this recording."
ChatGPT told me it would review the audio file, focusing only on the female voice and assessing it "like Vinh Giang would: focusing on clarity, vocal variety, pace, presence and confidence." (At least it knows who Vinh Giang is and understands the prompt!)
The AI tool further said it would extract and focus on 5 to 7 minutes from the 53-minute audio file and confidently said it would then be able to give me advice on my vocal tone and warmth, pace and pausing, clarity and articulation, confidence and conviction, presence, and energy. But, well, it didn't.
And when it couldn't, it told me I had also reached my data analysis limit on the free version of ChatGPT.
Since I was using the free version, I asked ChatGPT to critique the 1-2-minute portion first, before I upgraded to paid. It couldn't extract the 2-minute section, so I had to edit the file and reupload it, which was a pain. I quickly duplicated the file then trimmed it to two minutes, at a point where I was speaking.
Then, it would break down vocal variety, pacing and pausing, clarity and articulation, confidence, conviction and presence.
But it still couldn't.
While the transcript could be helpful, I wanted ChatGPT to review my actual voice, so I upgraded to ChatGPT Plus, which costs $20 a month.
I tried to upload the entire recording again, but it still had a hard time with it.
So, I re-uploaded the 2-minute clip. But I got cut off again, telling me: "Transcription tools aren't currently available in this environment."
That would've been helpful for ChatGPT to tell me before it took my $20.
I uploaded my entire transcript. Again, it could only review 1 to 2 minutes, and it pulled out the first section, which is the introduction -- not ideal to critique the early call pleasantries.
It was nice about me, though.
It did give me some sage advice too. I've got the connection side nailed, I just need to work on my confidence.
For more specific advice, I copied a part of the transcript where I talk a lot. I used the prompt: "Critique this section of the transcript," then copied the text, including the speaker titles.
Now we're talking:
ChatGPT said my strengths are being personal and approachable, but I use filler words a lot, can ramble and say sorry too much. Definitely true.
ChatGPT advised me to avoid language that seeks permission and rather speak in declaratives. It suggested daily identity affirmation practices and to train my voice like a musical instrument. It offered to build me a seven- or 30-day voice training routine, which I was curious to see.
Here's an example:
To truly understand AI's coaching capabilities, I tried a different AI tool that offers roleplays and reviews. With Yoodli AI, I uploaded that same audio clip to see what it could do.
After scanning the advice from the AI coach, I realized it had critiqued the interviewee, not me. There was no Yoodli chatbot to tell it to focus on the interviewer, so most of the advice was redundant.
This is where the role-play comes into play. I could have a video call with the AI tool and practice speaking.
I could even create a custom scenario, telling the tool what I want it to be and asking it to construct the conversation.
Very cool.
While I liked how AI confirmed some assumptions I'd had, it's not a replacement for proper training. It could be handy to use these tools to rehearse and get real-time feedback, but it won't turn me into a pro speaker anytime soon.
