How to Install and Use Whisper by OpenAI to Convert Video to Text

Step 1: Install Whisper First, you’ll need Python installed (version 3.8+ recommended). To install Whisper, open your terminal or command prompt and run (You can use anaconda for isolated environment):

pip install -U openai-whisper 

✅ Tip: If you don't have pip or Python, download Python here. Step 2: Install ffmpeg Whisper needs ffmpeg to handle video and audio files. Install it with: # For Ubuntu/Debian brew install ffmpeg # For Mac (with Homebrew) choco install ffmpeg # For Windows (with Chocolatey)

sudo apt install ffmpeg

Or you can manually download ffmpeg from ffmpeg.org. *Note the above link will redirect you to a github page and you can download appropriote one for your operating system. I selected: ffmpeg-master-latest-win64-gpl.zip

1. Download the binary edition (zip/rar archive) and extract the files into C:\ffmpeg
2. There should be three files named: ffmpeg.exe, ffplay.exe and ffprobe.exe
3. Then you must add the ffmpeg path (C:\ffmpeg\) to the system environment variables.
4. Search for "View advanced system settings" and open it.

5. And here you should select: System variables > Path > Edit > New > "C:\ffmpeg" > OK

Now you should be able to invoke ffmpeg on Command Prompt (cmd.exe). *Note: If you had already opened cmd.exe before adding ffmpeg, you should reopen cmd.exe to see the taken effect. Step 3: Convert Your Video to Text Now, you can easily convert any video file to text with just one command. Example:

whisper your_video.mp4 --model small 

Replace your_video.mp4 with your file name. Example result: The output will be your_video.txt file automatically created in the same folder! Bonus: Full Python Script to Use Whisper Programmatically Want to do it inside a Python script? Here’s a simple example:

import whisper
model = whisper.load_model("base") # You can also use "small", "medium", or "large"

result = model.transcribe("your_video.mp4")

print(result["text"])

✅ This script will print the full text transcription! All models and comparison among them are given below:

Model	Parameters	Size on Disk	Speed	Accuracy	Hardware Need	Typical Use Case
Tiny	39M	~151 MB	Very Fast	Lowest	Very Low	Real-time apps, mobile devices
Base	74M	~290 MB	Very Fast	Low-Mid	Low	Faster decoding with slightly better quality
Small	244M	~967 MB	Fast	Medium	Medium	Good quality at a reasonable speed
Medium	769M	~2.9 GB	Moderate	High	High	Higher accuracy needs, server-side apps
Large	1550M	~5.8 GB	Slowest	Best	Very High (Strong GPU recommended)	Best quality transcription, multilingual, research

Optional: Whisper Command Line Options Here are some handy extra options you might like: Option Description --language English Set the language manually --task translate Translate speech to English --output_format txt Save output as .txt --model small Use a smaller, faster model Example command:

whisper your_video.mp4 --language English --task translate --output_format txt --model small 

Input: lesson1.mp4 Outputs (all):

• lesson1.json (timing in json)
• lesson1.srt (with timing)
• lesson1.tsv (start end timestamp)
• lesson1.txt (Only text)
• lesson1.vtt (with timing)

Final Thoughts Whisper is super powerful and surprisingly easy to use once you set it up. You can automate transcriptions, translations, and even batch process tons of videos! Happy transcribing! ✨