sentdex
Artificial intelligence teaches you how to incorporate your deep learning models into user interfaces for humans, using Gradio.
Created using #copilot for the code, Riva (https://developer.nvidia.com/riva) from NVIDIA for custom speech synthesis, and a handful of LLMs for scripting.
Things are getting interesting.
Neural Networks from Scratch book: https://nnfs.io
Channel membership: https://www.youtube.com/channel/UCfzlCWGWYyIQ0aLC5w48gBQ/join
Discord: https://discord.gg/sentdex
Reddit: https://www.reddit.com/r/sentdex/
Support the content: https://pythonprogramming.net/support-donate/
Twitter: https://twitter.com/sentdex
Instagram: https://instagram.com/sentdex
Facebook: https://www.facebook.com/pythonprogramming.net/
Twitch: https://www.twitch.tv/sentdex
#programming #shorts #pythonprogramming
Source
Crazy how much improvement we've seen already in this speech synthesis model. Would be interesting to plot out quality over time and estimate how long it will be before the speech synthesis model can fool your average viewer!
the audio in this video was text-to-speech right?
Wait what wow 😮
Did you use an AI to make this video ? lol
why does it sound like a robot voiced this video
Very cool sentdex! I feel like I'm going to be replaced soon though 😱
Before I realized who's voice it is It sounded just like another robotic voice. But now that I know, it somehow become harder to tell that it is TtS. 😲
This is how I lost my job, honey.
model = keras.Model(inputs=inputs, outputs=outputs)
Output tensors of a Functional model must be the output of a TensorFlow `Layer` (thus holding past layer metadata). Found: <keras.layers.core.dense.Dense object at 0x7f772e256790>
This is the error im getting on this line
I just copy paste your code
this is amazing
Soon ai will teach human how to be a human
Presumably the vocoder trained on the sound of your voice, but didn't really have a dataset of the words you said?
Under the condition that this presumption is correct, I wonder if the synthesis quality would be improved with fully classified audio. i.e. supervised fine-tuning rather than just learning pitch/phonetic parameters
Amazing tool!
Hey your voice looks generated.
Can you write the name of the ai as the title the next time you show us ai
Nice voice
Your pixels are showing, Sentdex 🌝
For once I would have liked subtitles, the weird diction made it hard to understand for me. And I didn't understand what gradio is, need to rewatch😅