Discover Whisper, the best speech-to-text tool developed by OpenAI

Whisper speech to text

Hi company, it's your humble servant, Nicolas, from!

Ahoy, today is a day of celebration! I have the honor, nay, the privilege, of introducing Whisper, this prodigy of Speech-to-Text technology, the fruit of the incredible labor of our friends at OpenAI. You know, those geniuses who gave birth to power monsters like ChatGPT, GPT-3, 3.5, 4 and all the rest of the family… The little story goes that they sold their souls to science, but that's another story!

A transcription more accurate than a Swiss watch!

So, what's in store for our new friend Whisper? Well, for a start, this little gem has already been around for over a year and a half. And what a feat! It provides surgically precise transcriptions for all your linguistic needs: English, French, Italian, Spanish… It can do it all. Whether you need a transcription of a YouTube video or an audio recording on your smartphone in the blink of an eye, Whisper takes care of it all. And best of all, it translates it into English as a bonus. Perfect for impressing your buddies at the aperitif


Let's take off for the wonderful world of Whisper!

Come on, enough blabla, let's take a closer look at our beautiful discovery. Whisper's presentation page is a veritable catalog of technological prowess: transcription of faster-than-light speech, French content (but yes, you know, that language with lots of incomprehensible rules), K-Pop videos (ideal for learning to dance at the same time), and even spoken words with an accent! Hats off to you, Whisper.

For the tech-savvy who want to dig under the hood, Whisper offers a section dedicated to its internal engine. Tokens, encoding, decoding, it's all there. It almost sounds like a course in quantum mechanics, but rest assured, you don't need to be Albert Einstein to use the basic tool


Instructions for using Whisper

Ready, set, go! Let's set off on an exciting journey into the heart of Whisper usage. First of all, don't panic, access is free on Google Collab, with no restrictions. Yes, you read that right, free. So how do you do it?

Here's the link to the tool:

Step 1:GPU access is checked on Google Collab.

First stop, the GPU on Google Collab. Why? To give wings to our transcription. To check, just click on “Modify execution type” in the top right-hand corner of your Collab page. Check that you're on the “T4” GPU and type, the Ferrari of free GPUs


Step 2: We install the necessary libraries.

Second stop, the Python libraries. Don't panic, a little script in the first cell does all the work for you. It's like having a personal cook who prepares everything while you enjoy your aperitif


Step 3: We configure the backup folder.

Step three, define where Whisper will store all those precious transcripts. Google Drive or another local folder, the choice is yours. Don't worry if the folder doesn't exist, Whisper will create it for you. Isn't that nice?

Step 4: We choose our model.

Step four, choosing your model. A crucial choice, a bit like choosing your ice cream flavor. Whisper offers a beautiful palette, from “tiny” to “large”. The “medium” is often a good compromise between speed and precision


Step 5: The video to be transcribed is selected.

Fifth stop, choose the video to transcribe. A YouTube video or a local file, it's up to you. A simple copy and paste for YouTube, or a selection of the local file, and you're done.

Step 6: Output options are configured.

Sixth stop, the output options. Do you want a plain text or structured format like JSON, VTT, SRT, TSV? Whisper adapts to your wishes

Step 7: We run the model.

Seventh step, we turn on the turbo. Click on the button to launch the cell, and Whisper goes to work. A little patience and you've got your transcription ready to go


Step 8: The transcript is analyzed and checked.

Eighth and final step, a quick look at the transcription. Whisper is a pro, but nobody's perfect. A few small manual corrections may be necessary.

Whisper, our everyday friend

Whisper is disconcertingly simple. No need to know Python or tinker with code. Just select, click and you're done. You'll get a transcription as precise as a Swiss clock, ready to go


A tool with a thousand facets

Whisper is like a Swiss Army knife, it's got lots of uses. Want to transcribe YouTube videos into different languages, translate audio content, take notes in meetings or classes? It's there for you. Whisper is the ideal companion for anyone who needs fast, accurate transcriptions. And all while sipping your coffee. What more could you ask for?

Leave a Reply

Your email address will not be published. Required fields are marked *