Every journalist, copywriter, and blogger knows how tedious it is to process interviews. Transcribing recordings and editing text can take anywhere from a few hours to an entire workday. Artificial intelligence services can help simplify the process and minimize the time required.
These instructions are suitable not only for creating interviews, but also for all cases where you need to quickly turn audio or video into understandable and readable text. For example, you can repurpose webinars, conferences, lectures, podcasts, or any YouTube video into articles for blogs or the media, press releases, and social media posts.
As an example, we will turn a 1.5-hour interview into text suitable for publication and see how long it takes.
Step 1. Convert speech to text using transcription software
Speech recognition services or transcription software can help convert audio or video files into text: Speech2Text, Any2Text, Teamlogs, and others.
Copy the link to the YouTube interview and paste it into the appropriate window on the service; in our case, we chose Speech2Text. This program can break text into paragraphs, add punctuation marks, and recognize speakers by their voices, which is very convenient for transcribing interviews.
You can also upload audio and video files in any format to the transcription program.
Insert the link to the video on the Speech2Text homepage.
The service will determine the language itself, and the number of speakers can also be left for automatic recognition. After clicking the “Recognize” button, the service will indicate the estimated processing time for the file. It took only 17 minutes to process a 90-minute video.
Next, open the interactive player and determine which lines belong to each of the two speakers by checking the timecodes. Return to the main page and enter the names of “Speaker 1” and “Speaker 2” in order of appearance.
By the way, when editing the text in the interactive player, it is convenient to listen to unclear parts of the transcription and double-check terms and names. To do this, click on the timecode, and the video or audio file will start at the right place.
Everything is ready, now all that remains is to download the file in docx format to your computer. The transcript can be obtained with or without timecodes, in the following formats:
- continuous text;
- text divided by speakers;
- text divided by speakers and paragraphs.
To prepare the interview for publication, the best option is text divided into paragraphs and speakers, without timecodes. The file downloads in a minute.
In total, we spent 3 minutes copying the link, specifying the names of the speakers, and downloading the file. We do not count the 17 minutes spent transcribing the video, as we were able to do other things during that time.
Step 2. Correct errors and style the text using a neural network
AI can tidy up text: correct errors, shorten it, and even improve its style. For example, you can use Online NotePad, Service-online, or another similar service. Select a piece of text in the online editor, right-click, and then select “Help with text” from the pop-up menu.
You can use all three functions in turn. A minor inconvenience is that you will have to process the text in pieces. The neural network will refuse to work if you select the entire text at once.
Correcting errors
After selecting the “Correct errors” function, the neural network added missing commas, colons, and other punctuation marks and corrected errors.
Shortening the text
The AI offered several options for shortening the text, with varying degrees of conciseness. After applying the “Shorten” function, the text was reduced by half, and the sentences became shorter and clearer. Filler words such as “well,” “here,” and ‘some’ were removed, and the incorrect title of the book “Describe-Shorten” was corrected to the correct “Write, Shorten.”
Improving style
The “Improve Style” function helped to remove colloquialisms and conversational expressions. For example, the expression “I’m finishing the pictures” was replaced with “I’m finishing the illustrations,” “everything went a little wrong” with “everything went a little off plan,” and “That’s clear and understandable” was replaced with “That’s what ‘clear and understandable’ means.” As with the abbreviation, several options for improving the style were suggested.
Time spent improving the style of the text: 20 minutes.
Step 3. Edit the text manually
In step 3, you need to manually go through the text and remove any remaining stylistic errors, change individual expressions to your liking, and check names and terms.
