Free local transcription in Whisper Web works on files up to 20 minutes and 200 MB. If your interview, lecture, or meeting is longer, you do not have to give up. You can split it into parts, transcribe each part, and join the text. This guide shows how, in plain steps.

The free path runs in your browser, so the audio stays on your device. Longer files, batch uploads, and saved history are part of Pro, which uses cloud processing, so the media is uploaded. Splitting is for people who want to stay on the free local path.

Quick answer

Check both limits. The free local limit is 20 minutes and 200 MB per file. A file must fit under both.
Split at pauses. Cut where the speaker stops, not in the middle of a sentence.
Name files in order. Use 01, 02, 03 so you never mix up the parts.
Join and check. Paste the parts together and read the places where they meet.

Step 1: Work out how many parts you need

Take the total length and divide it by 20, then round up. A 55-minute interview needs three parts. Aim for equal parts of about 18 minutes. That leaves a safety margin if a part runs a little long.

Do not forget the size limit. File size depends on the format. A 55-minute MP3 at 128 kbit/s is about 53 MB, so size is rarely a problem. A CD-quality stereo WAV is much bigger. One 18-minute part is about 190 MB, which is very close to the 200 MB limit. If a part is too big, make the parts shorter or save the audio as MP3 first.

Step 2: Find natural break points

Open the recording in any audio editor you already use. Many free ones can split a file. Look for the end of a question, a change of topic, or a long pause. These are safe places to cut.

Here is an example. You have a 55-minute interview. Around minute 18, the interviewer finishes a question and takes a breath. Cut there. Around minute 37, the guest finishes a story and there is a short silence. Cut there. Now you have three parts, and none of them starts in the middle of a word.

What to avoid

Do not cut in the middle of a word or number.
Do not cut while two people are talking at once.
Do not rename parts by hand later. Name them correctly from the start.

Step 3: Name and keep the parts in order

Use a simple pattern such as interview-01.mp3, interview-02.mp3, and interview-03.mp3. Keep the original long file in a safe place. If something goes wrong, you can always split it again.

Step 4: Transcribe each part

Open Whisper Web audio to text, drop in the first part, pick the language or let it detect the language, and start. The first local run may take longer while the speech model downloads, so keep the tab open. After that, run the other parts the same way.

Download each result as TXT or Word, and name the text files the same way as the audio: 01, 02, 03. If you also need subtitles, you can export SRT or VTT for each part.

Step 5: Join the text and fix the seams

Paste the three texts into one document in order. Then read the places where the parts meet. The seam is where mistakes hide. A sentence may be cut, a word may be repeated, or a name may be spelled differently in each part.

A simple example: in part one the guest is written as "Saoirse", and in part two as "Sear-sha". Search for the wrong spelling and replace it everywhere. Do the same for brand names and numbers.

A note about subtitle timing

Each part is its own file, so its timestamps start at zero. If you join SRT or VTT files, you must add the length of the earlier parts to every time in the later parts. For a video you plan to publish with subtitles, it may be simpler to use a tool that handles the whole file, such as Pro.

Step 6: Check names, numbers, and quotes

Accuracy depends on the language, the clarity of the speech, and the quality of the recording. Accents, background noise, and people talking over each other cause errors. Before you rely on the text, check names, numbers, and any quote you plan to use against the audio.

When splitting is not worth it

Splitting is great for one or two long files. It gets tiring when you have many hours of audio. In that case, Pro supports longer files and batch uploads, with cloud processing. Choose it when you are comfortable with the media being uploaded. If the recording is confidential and you cannot upload it, stay with the free local path and split.

A short checklist

Divide the length by 20 and round up to find the number of parts.
Check that every part is under 200 MB.
Cut at pauses and name files 01, 02, 03.
Transcribe each part and download TXT or Word.
Join the text and read every seam.
Check names, numbers, and quotes against the audio.

Ready to try it? Open Whisper Web audio to text and start with the first part. For more on keeping recordings on your own device, see how to transcribe audio without uploading it.

Transcribe a Long Recording for Free: How to Split It Under the 20-Minute Limit