A podcast transcript is one of the highest-leverage assets a show can produce: it feeds SEO, accessibility, show notes, and repurposing. The good news is that you no longer need a subscription to get one. OpenAI's Whisper model — the engine behind most modern transcription tools — is free and open source, and several services give away genuinely usable free tiers built on top of it.
Below are three complete workflows that cost nothing: a quick cloud option, a fully unlimited local option for Mac users, and a free Google Colab route that works on any computer.
Option 1: TurboScribe (easiest cloud route)
TurboScribe is a browser-based transcription service built on Whisper that supports 98+ languages. Its free tier is unusually generous for occasional work: 3 transcriptions per day, with each file up to 30 minutes long, and no credit card required. Unlimited costs $20/month billed monthly or $120/year if you ever outgrow the free plan.
This is the right pick if you want a transcript in a few minutes with zero setup.
Steps
1. Sign up at TurboScribe with Google or email — the free tier activates immediately.
2. Click upload and add your episode file (MP3, WAV, M4A, and video formats all work). Keep each file under 30 minutes on the free plan.
3. Select the audio's language from the dropdown. Auto-detect works, but picking the language explicitly tends to produce cleaner results.
4. Leave the transcription mode on default and start the upload. A 30-minute episode typically processes in under three minutes.
5. Review the transcript in the built-in editor. TurboScribe highlights low-confidence words — click through them and correct against the audio.
6. Export in your format of choice: TXT for show notes, DOCX for editing, or SRT if you need captions for a video version.
One limitation to know: on the free tier, speaker labels and AI summaries sit behind the paid plan, and a long episode will be chopped at the 30-minute mark. For interviews where you need to know who said what, export the raw text and add speaker labels manually, or use one of the local options below.
Option 2: Whisper on Google Colab (free GPU, works anywhere)
If you have more than a couple of episodes to process, the free tiers run out fast. The unlimited answer is Whisper itself, and you do not need a powerful computer: Google Colab gives you a free GPU runtime in the browser.
This workflow sounds technical but takes about five minutes to set up, and every step is copy-paste.
Steps
1. Go to Google Colab and open a new notebook. Under the Runtime menu, change the runtime type to a GPU — transcription runs dramatically faster with one.
2. In the first cell, install Whisper and ffmpeg:
!pip install -q openai-whisper
!apt-get install -y -q ffmpeg3. Upload your episode: use the file browser pane on the left and drag the MP3 in.
4. In a new cell, pick a model and transcribe:
!whisper "episode.mp3" --model small --language en --output_format srt Model choice is a speed-versus-accuracy trade-off: tiny is fastest, base and small are the sweet spot for podcasts, and large is the most accurate but noticeably slower.
5. Download the resulting episode.srt, episode.vtt, or episode.txt from the file pane before closing the tab — Colab discards everything when the session ends.
6. Skim the transcript once against the audio at 1.5x speed. Proper nouns, brand names, and overlapping speech are where Whisper slips most often; fix those by hand.
There is no daily limit beyond Colab's fair-usage GPU time, so this route handles entire back catalogs at zero cost. If a podcast's episodes are longer than an hour, split them into chunks first to avoid session timeouts.
Option 3: MacWhisper (free local app for Mac)
Mac users get a particularly clean free option: MacWhisper, a free app that runs Whisper entirely on your machine. Nothing is uploaded, which matters if your episodes contain unreleased material or sensitive interviews. Apple silicon Macs are recommended, especially for the newer models.
Steps
1. Download MacWhisper from the official site and launch it.
2. On first run, pick a model. Small or Medium balances accuracy against speed on most Apple silicon machines.
3. Drag an episode onto the app window. It transcribes locally, file by file, with no queue limits and no per-day caps.
4. Use the built-in editor to clean up the transcript — it includes speaker recognition and subtitle export in the free version.
5. Export as TXT, SRT, or subtitles for your video edit.
If you want more power later, MacWhisper Pro is a one-time license (€64, lifetime updates) rather than a subscription. For open-source purists, no-typing-mac is a free community-built alternative that also runs Whisper locally on macOS with Metal GPU acceleration.
What about Notta and Otter?
Two names come up constantly, but both are built for live meetings rather than podcast archives. Notta has a free plan and paid tiers starting at $8.17/month, with real-time transcription in 58+ languages — handy if you transcribe while recording, less so for a backlog of MP3s. Otter's free plan gives 300 minutes a month with a 30-minute session cap and only three lifetime file imports, which disqualifies it for archive work almost immediately. For podcasts, a file-upload tool like TurboScribe or a local Whisper setup is simply the better shape of tool.
Getting cleaner transcripts
The model matters less than the audio. A few habits pay off no matter which workflow you use:
- Record in WAV or high-bitrate MP3. Heavily compressed audio at 64 kbps is the single biggest cause of garbled transcripts.
- Give each speaker their own track when possible. Two people on one track with crosstalk is where every free tool struggles.
- Reduce room noise before uploading. Most tools do a passable job with mild hiss, but HVAC hum and keyboard clatter cost you real accuracy.
- Do a 60-second test first. Run the intro of an episode through your chosen tool before committing a 90-minute file — it reveals pronunciation issues with names and jargon early.
- Budget a cleanup pass. Whisper is excellent, but names, numbers, and quotations still need a human eye. Listen at 1.5x while reading, fix the errors, and the transcript is publish-ready.
Turn the transcript into more content
The transcript is not the end of the pipeline — it is raw material:
- Paste it into your show-notes draft and pull out the three strongest quotes for social clips.
- Search engines index text, not audio. Publishing the transcript (or a cleaned summary of it) on your episode page gives each episode a real chance to rank for long-tail keywords.
- Feed the transcript to a free LLM and ask for a 150-word episode summary, five key takeaways, and ten chapter titles — instant episode packaging.