5 reasons NOT to use ChatGPT to transcribe videos

You cannot directly use ChatGPT to transcribe videos, but you can use OpenAI’s Whisper API to transcribe the audio of your video file.

But I would recommend using AssemblyAI instead to transcribe your videos.

Here is why:

It is much easier than ChatGPT if you need to edit your transcript

Since you call AssemblyAI via API, you get the response in JSON format, which means you can write some code to process this JSON and repurpose it into articles, summaries etc.

Here is an example JSON response for this YouTube video.

The JSON can be parsed to identify speaker labels

The JSON output from AssemblyAI provides speaker labels out of the box, so you can include speaker names in your transcript if you would like.

Note: Whisper API does not provide speaker labels out of the box

You can get Expanded Chapters

You can also ask AssemblyAI to add “auto chapters”. This provides three fields – a heading, a gist and a summary

Please note that these auto-chapters are generated by a Large Language Model (LLM) and as with other LLMs, it might sometime hallucinate information which isn’t present in the video.

So you need to proofread the output before publishing it.

But you can combine these fields and create a timestamped summary of your video.

The JSON format is easier to repurpose for multiple use cases

The JSON format is also easier to repurpose for multiple use cases.

For example you can write some code to turn the response into an article-like format which includes subheadings.

Here is an example article based on the JSON response above.

Reliable and consistent output

The heading, gist and summary generated by AssemblyAI are consistent.

In other words, for the same input audio file, you get the same automatic chapters.

On the other hand, if you used ChatGPT API for formatting or summarizing your video transcript, the results will differ each time you call the API.

If you prefer consistent summaries, AssemblyAI is a better choice.