Free Transcription Tools Compared (as of October 2026)
Updated
Only two options in this comparison transcribe with no minute limit, no sign-up and nothing to install: TranscriptBalance, which runs in your browser on your own device, and Riverside's free tool, which processes your file on its servers. The rest either cap free use (Otter, Notta, Happy Scribe, TurboScribe), have to be installed (aTrain, noScribe) or need a Microsoft 365 subscription (Word).
TranscriptBalance is our own product. Everything said about the other services comes from their official pages and was checked on 9 October 2026; the sources are listed below the table. We did not measure accuracy. Free plans change often, so check with the provider before you decide.
Free transcription tools at a glance
| Tool | Free | Account needed | Where the audio is processed | Live transcription | Languages and notes |
|---|---|---|---|---|---|
| TranscriptBalance | Unlimited: no minute, file or daily cap; only your device's memory sets a limit | No | On your device, in the browser (WebGPU or WebAssembly); only the models are downloaded, once | Yes, with no time limit: audio from a tab, a window or the whole screen (Chrome, Edge on a computer) or the microphone; plus AI questions | 99 languages (Whisper); speaker detection for files; export to TXT, DOCX, SRT, VTT, MD, JSON, CSV; speed depends on your device |
| Otter (Basic) | 300 min per month, up to 30 min per conversation, 3 file imports in total (not per month) | Yes | In the provider's cloud | Yes, also as an assistant in Zoom, Microsoft Teams and Google Meet (one meeting at a time) | English, Spanish, French, German, Japanese, Chinese; speaker identification; iOS and Android apps |
| Notta (Free) | 120 min per month, up to 3 min per conversation, 50 file uploads per month | Yes | In the provider's cloud | Yes, including meetings in Zoom, Google Meet, Teams and Webex; on the free plan only up to 3 min per conversation | 58 languages; 10 AI summaries per month |
| Happy Scribe (Free) | 10-minute trial of AI transcription; unlimited meeting recordings of up to 45 min each | Yes | In the provider's cloud | Meeting bot for Google Meet, Zoom and Teams; for details see the provider | 60+ languages and dialects on the free plan (the provider's comparison table says 150+); MP4 exports are watermarked |
| Riverside (free tool) | Unlimited, according to the provider | No | Uploaded and processed on the provider's servers | No, files only (MP3, WAV, MP4, MOV) | 100+ languages; no speaker separation; export to TXT and SRT |
| Microsoft Word (Dictate, Transcribe) | Only with a Microsoft 365 subscription; Transcribe then covers up to 300 min of uploaded recordings per month | Yes, with a Microsoft 365 subscription | By Microsoft (internet required); recordings for Transcribe are stored in OneDrive | Dictate: yes, from the microphone straight into the document. Transcribe: text only after the recording | Transcribe separates speakers and supports 80+ locales, on the web only in Edge and Chrome; Dutch is a preview language for Dictate |
| TurboScribe (Free) | According to the provider, 3 files per day of up to 30 min each, one file at a time | See the provider | Uploaded and processed by the provider | See the provider | According to the provider, 98+ languages, based on Whisper; free files wait in a lower-priority queue |
| aTrain | Free and open source (AGPL-3.0), no quota | No, but it has to be installed | On your computer, offline | Recorded files only | 99 languages; speaker detection; files for MAXQDA, ATLAS.ti and NVivo; Windows, Linux, macOS (CPU only there); made at the University of Graz |
| noScribe | Free and open source (GPL-3.0), no quota | No, but it has to be installed (installer about 3 GB) | On your computer, offline | Recorded files only | Around 60 languages; speaker detection; built-in editor for corrections; according to the provider 1 to 3 hours of processing per hour of interview; Windows, macOS, Linux |
Which tool fits which need?
- No upload, no account, no limit: TranscriptBalance runs in your browser with nothing to install. How that works without uploading anything is explained in Private transcription without uploads.
- Old or slow device, long file, little time: a cloud service runs on its own servers and is often faster than your laptop, for example Riverside's free tool (no account) or TurboScribe. In return, your recording goes to the provider.
- A meeting bot in Zoom, Teams or Meet, a phone app, transcripts on all your devices: that is what Otter, Notta and Happy Scribe are built for, and Otter has apps for iOS and Android. On the free plans, though, each conversation stops after 30 minutes (Otter), 3 minutes (Notta) or 45 minutes of recording (Happy Scribe).
- You already have Microsoft 365 and write in Word anyway: Dictate and Transcribe are built in, with no extra software.
- Interviews for your thesis or dissertation, analysed in MAXQDA, ATLAS.ti or NVivo: aTrain produces files for those programs, and noScribe has an editor for corrections. Both run locally but have to be installed. TranscriptBalance needs no installation and exports with speakers and timestamps, for example as DOCX or CSV. More on this: Transcribe interviews for free.
- Taking notes live in lectures, seminars or Teams meetings: TranscriptBalance's live mode has no time limit and its AI answers your questions about what was said. More on this: Live lecture transcription.
Where does TranscriptBalance fall short?
- Speed depends on your device. With WebGPU the graphics card does the work; otherwise Whisper runs on the processor through WebAssembly, which takes much longer. The most accurate model (Precise) needs a GPU; on weaker devices the app suggests a smaller model.
- A one-time download. The first time, your browser downloads the model you picked, about 45 to 760 MB depending on the model and device. After that it stays in the browser's storage.
- The tab has to stay open while it transcribes.
- No account also means no cloud. Results are kept only in this browser, with no sync between devices and no team sharing. Export what you want to keep.
- No app, no meeting bot. Sharing the audio of a tab or the screen only works in Chrome or Edge on a computer. In other browsers and on phones or tablets, live mode listens through the microphone.
- Live mode uses smaller models. To keep up in real time it runs smaller models than file mode (a dedicated model for German), and it does not tell speakers apart.
- AI answers need an AI. In live mode the answers come either from Chrome's built-in AI (a one-time download of about 4 GB, not available on every device) or from Gemini with your own Google key (18+; Google’s terms and limits apply). With Gemini, text excerpts go straight to Google, never your audio.
- Mistakes happen. Like every automatic transcription, Whisper mishears words, especially technical terms, strong accents and poor audio. Check the text before you use it.
The quickest way to find out is to try a file of your own: no account, and your audio stays on your device.
Open TranscriptBalance for freeSources
All details as of 9 October 2026.
- Otter: Pricing and plans, Help Center: languages
- Notta: Pricing and plans, Languages
- Happy Scribe: Pricing and plans
- Riverside: Free transcription tool
- Microsoft: Transcribe your recordings, Dictate your documents in Word
- TurboScribe: Pricing (details as stated on the provider's website)
- aTrain: Project page
- noScribe: Website, Project page
- TranscriptBalance: Privacy policy (what leaves your device and what does not)
Frequently asked questions
Which transcription tool is really free and unlimited?
In this comparison from 9 October 2026, TranscriptBalance, aTrain and noScribe have no quota, because they run on your own device; TranscriptBalance does so in the browser with nothing to install. Riverside's free tool is also unlimited according to the provider, but it processes your files on its servers. Otter, Notta, Happy Scribe and TurboScribe cap free use, and Word only transcribes with a Microsoft 365 subscription.
Which free tools transcribe without uploading my recording?
TranscriptBalance (in the browser) and the desktop programs aTrain and noScribe process the audio on your device. TranscriptBalance only downloads its AI models once, from Hugging Face and jsDelivr; your audio is not uploaded. Otter, Notta, Happy Scribe, Riverside, TurboScribe and Word process the recording on the provider's side.
Can I transcribe lectures live for free?
Yes. TranscriptBalance's live mode transcribes with no time limit: in Chrome or Edge on a computer it takes the audio of a tab or the screen, and in other browsers and on phones it uses the microphone. On their free plans, Otter transcribes up to 30 minutes per conversation and Notta 3 minutes. Let your lecturer know before you transcribe.
Is there a free, unlimited alternative to Otter or TurboScribe?
Yes, if you can do without a meeting bot and a cloud account. TranscriptBalance (in the browser) and aTrain and noScribe (to install) transcribe on your device with no quota, and Riverside's free tool is unlimited on its servers according to the provider. What these alternatives do not offer: a bot that records meetings automatically, phone apps, and transcripts in an account on all your devices.
Why is TranscriptBalance free, and what's the catch?
Your own device does the computing, so a minute of transcription costs us no server time. TranscriptBalance has no subscription and no account; Jeremy Heindrichs, a student in Belgium and the developer of the Windows app AudioBalance, builds it alongside his studies. The limits: speed depends on your device, a model is downloaded the first time, and there are no cloud or team features.
How accurate is free transcription?
It depends mostly on the audio quality, the language and the model; we did not measure accuracy for this comparison. TranscriptBalance uses Whisper, and you choose between four models from Flash to Precise: larger models are more accurate but need more computing power. Check every automatic transcript before you quote from it or hand it in.