How to Transcribe a Long Audio Recording (1 to 5 Hours)
When you need to split a long recording and when you don't, the limits of other tools, how to shrink the file with ffmpeg, and how to proofread a multi-hour transcript.
Sonorarium Team 5 min read
Short answer: if your recording is up to 5 hours and 2 GB, upload it whole, without splitting. If it's bigger, the cause is almost always the format (uncompressed WAV or a video file): convert it to compressed audio and it shrinks to about 29 MB per hour. If it runs longer than 5 hours, split it into two or more pieces. Then don't read the transcript start to finish: search for names and figures, and use timestamps to listen only to the doubtful passages.
This guide is for files of 1 to 5 hours of any kind: long interviews, hearings, conference days, work sessions. If you have a 90-minute class and want study notes, there's a separate guide for lectures.
Do you need to split the file?
It depends on each tool's limit. These are the ones published on the official pages, checked on October 26, 2026 (they can change):
| Tool | Published limit | Source |
|---|---|---|
| Otter | Per conversation: 30 min (free), 90 min (Pro), 4 h (Business). The free plan allows 3 file imports in total | otter.ai |
| Happy Scribe | Per recording: 45 min (free), 90 min (Basic), unlimited (Pro and Business) | happyscribe.com |
| Notta | Per recording: 3 min (free), 5 h (Pro and Business) | notta.ai |
| OpenAI API (for developers) | 25 MB per file; it recommends compressing or splitting | docs |
| Word (Microsoft 365) | 300 minutes of uploaded audio per month | Microsoft |
| Sonorarium | 2 GB and 5 hours per file | Our own limit |
With a 30- or 90-minute cap, a 3-hour recording becomes 2 to 6 files. Splitting has a price: OpenAI's own documentation warns that cutting mid-sentence can remove context and reduce accuracy, you have to stitch the texts together afterwards, and the timestamps restart at zero in every piece.
With Sonorarium you only split in two cases: the file is over 2 GB and you can't shrink it, or it runs longer than 5 hours.
How to shrink the file
A big file is usually big because of its format, not its length. Approximate sizes per hour of recording (our own calculation from the bitrate):
| Format | Size per hour |
|---|---|
| WAV stereo, 44.1 kHz, 16-bit | about 635 MB |
| MP3 stereo at 128 kbps | about 58 MB |
| MP3 mono at 64 kbps | about 29 MB |
A WAV like that reaches 2 GB at about 3 hours 20 minutes, so a 4-hour WAV doesn't fit. As mono MP3 at 64 kbps, 5 hours take about 144 MB.
For spoken voice, mono at 64 kbps is usually enough. If the original audio is already poor, compressing it won't improve it, so don't change more than you need to.
With ffmpeg, a free command-line program, these commands work (we tested them with version 9.0):
ffmpeg -i recording.wav -vn -ac 1 -b:a 64k recording.mp3
That converts to mono MP3 at 64 kbps. If you have a video that's too heavy, extract only the sound:
ffmpeg -i video.mp4 -vn -ac 1 -b:a 64k audio.mp3
Sonorarium also accepts video files (MP4, MOV, MKV and others), but the 2 GB limit counts the whole video, and a multi-hour video easily passes 2 GB even when its sound is small. If you prefer a program with windows, Audacity (free) can export to MP3.
If it runs longer than 5 hours
Split it into pieces under 5 hours. With ffmpeg, without re-compressing:
ffmpeg -i recording.mp3 -f segment -segment_time 14400 -c copy part_%03d.mp3
That cuts every 4 hours (14,400 seconds). If you can choose where to cut, a pause or a change of topic is better, using these two commands:
ffmpeg -i recording.mp3 -t 04:00:00 -c copy part1.mp3
ffmpeg -i recording.mp3 -ss 04:00:00 -c copy part2.mp3
When you join the texts, the timestamps of part 2 start at 0:00: add 4 hours so they match the original recording.
Before you pay: the free preview
If it runs 5 hours or less, upload the whole file without cutting it. The first 15 minutes are free when you create an account, with no card: if the file is longer, Sonorarium transcribes those 15 minutes and shows you the preview before you pay anything, so you check the language and quality on your own recording. If you like it, you unlock the rest from your account, without uploading the file again, and you only pay for the minutes that are missing. The file is kept for 7 days, so unlock it within that window. If the result is weak because of noise, see how to transcribe noisy audio.
On cost: minutes are bought in packs and don't expire. With the 300-minute pack ($14), a 3-hour recording uses 180 and leaves you 120 for the next one. There's a comparison with other options in how much it costs to transcribe one hour of audio, and the packs are on the pricing page.
How to proofread a multi-hour transcript
Four hours of audio is over 30,000 words. You won't read all of it carefully, so use a method:
- Search, don't read. The "Search the text" box highlights every match and dims the other lines. Search for proper names, acronyms, figures and key terms, and try variants (singular and plural, different spellings).
- Use the timestamps. They're on by default: each sentence carries its minute, so you can cite it and go back to the recording. If you don't see the minutes, check the "Timestamps" switch.
- Listen only to what's doubtful. Click a sentence's timestamp to hear it in the player; you can go back 5 seconds or raise the speed to review faster. Proofread within the 7 days after uploading: after that the audio is deleted and you can no longer listen to it.
- Fix what matters: figures, dates, names and verbatim quotes. The rest can stay as it is if the text is for reference.
- Add headings at topic changes, using the minutes as anchors.
One limitation: Sonorarium doesn't separate speakers. In a recording with several voices you'll have to mark who is speaking while you proofread; the timestamps tell you where.
When you're done, you can download Word, TXT, SRT, VTT or JSON; downloading and editing need a free account. The Word file puts a mark like [00:41:12] at the start of each paragraph, handy for searching later. And remember that the audio is deleted after 7 days, while the transcript in your account stays until you delete it.
The workflow in short
- Check the size: over 2 GB, convert to mono MP3.
- Check the length: over 5 hours, split it.
- Upload the whole file.
- Check the free preview and, if you like it, unlock the rest.
- Proofread by search and timestamps.
For long meetings, try meeting transcription; for interviews, interview transcription.