Every "100% free" transcriber has a catch somewhere — a monthly minute cap, a one-time trial, an account, or a paywall on the export. We compared eleven by name and wrote down what each one actually allows. Yes, ScribeGrab is ours; the limits below come straight off each tool's own pricing page, so you can check every line yourself. Free transcription apps, software you install, online services and old-style freeware are all in it, ranked by their actual limits. Most people arrive with a video, and that works in every one of them: the tool pulls the audio track out of the video file first, then transcribes that.
No account. 10 recordings a day. We just tell you the number. Files up to 60 minutes here, and it resets at midnight.
By ShinobiTools Team · Last updated: 2 October 2026
Checked on each tool's own pricing or product page on 2 October 2026, with the source in the last column. Free plans change often, so treat this as a snapshot with a date on it rather than gospel — where a limit wasn't published anywhere we could find, the table says so instead of guessing a number. Only need the best free audio transcription software for a pile of MP3s? The Free limit column is the one to read.
| Tool | Best for | Free limit | Account? | Watermark? | Price to go further | Source (checked 2 Oct 2026) |
|---|---|---|---|---|---|---|
| ScribeGrab (ours) | Long files, several a day | 60 min / 1.5 GB per file, 10 a day | No | No | €5 one-off for 30 days (or €3 for 24 hours) | our own code, see pricing |
| Otter.ai | Live meeting capture | 300 min/month, 30 min per conversation, 3 file imports per lifetime | Yes | Not mentioned | Pro from $8.33/user/month billed yearly ($16.99 monthly) | otter.ai/pricing |
| Whisper | Unlimited use if you're technical | No cap: you run it yourself | No | No | Free (MIT licence) | GitHub |
| MacWhisper | Mac users, files that never leave the Mac | Free version: local transcription, SRT/VTT, DOCX and PDF export | No | No | Pro €65 one-time (personal licence) | Gumroad page |
| Notta | Short clips | 120 min/month, 3 min per conversation, 50 file uploads/month | Yes | Not mentioned | Pro $8.17/month billed yearly ($13.99 monthly) | notta.ai/pricing |
| Descript | Editing audio/video like text | 60 media minutes/month, 100 one-time AI credits, 720p export | Yes | Yes | Hobbyist from $16/month | descript.com/pricing |
| Rev | Testing a paid service once | 45 AI minutes/month, English only | Yes | Not mentioned | Essentials from $25.49/seat/month; human transcription $1.99/min | rev.com/pricing |
| Happy Scribe | Trying the accuracy once | 10-minute one-time trial for files; meeting recordings up to 45 min | Yes | Yes (MP4 exports) | Basic €8.50/month billed yearly (€17 monthly) | happyscribe.com/pricing |
| Sonix | Trying the editor once | 30-minute trial, no card | Yes | Not mentioned | $10 per hour pay-as-you-go, or Core $25/month | sonix.ai/pricing |
| TurboScribe | A few short files daily | 3 files/day, 30 min each (could not re-check, see below) | Yes | Not checked | Unlimited $10/month billed yearly ($120/year) | turboscribe.ai/pricing blocked our check; figures from its search listing |
| Deepgram | Developers building their own tool | $200 free API credit, not minutes | Yes | N/A (API) | Pay-as-you-go after the credit | deepgram.com/pricing |
What each one is actually for, and where its free tier runs out. Numbers below are what each tool publishes on its own pricing page; where a company doesn't publish a ceiling, we say so instead of estimating one. All of them take video as well as audio: point the tool at the video file, a screen recording or a camera clip, and it pulls the speech out. ScribeGrab's free video transcription follows the same 60-minute, 1.5 GB limit as any audio file, with no account and no card. Whether you need an audio file transcriber or want to convert video to transcript and extract transcript from video, the steps are the same: upload, wait, download. Audio transcribing on a hosted tool does mean the audio file goes to their server first, which is the trade-off against the tools you run on your own computer. Speech AI in all of these has become good enough that the limits, more than the accuracy, usually decide.
Speech recognition isn't cheap to run: turning an hour of audio into text takes real GPU or CPU time, and someone pays for that per minute, whether it's rented from a cloud API or run on hardware the company already owns. That's the whole reason every "free" tier above has a limit somewhere — the limit is just where each company decided to draw the line.
None of the eleven tools above found a way around paying for GPU time — they just chose different places to put the bill.
There is no single winner here, because the tools solve different problems. Pick on the constraint that actually bites you:
We opened every tool's own pricing or product page on 2 October 2026 and copied the numbers it published there — no vendor emails, no press kits. Every row links to the page we read. Where a page doesn't mention something (Otter's, Notta's, Rev's and Sonix's watermark policy, for instance), the table says "not mentioned" rather than guessing, and TurboScribe's pricing page refused our automated check, which the table says too. Still working from a "best free transcription software 2025" list? The free transcription limits have moved since: this check is from October 2026, and we date every update so the page doesn't quietly go stale again.
For ScribeGrab we can be more precise, because it is ours. We measured our own speech model on 60 fragments of LibriSpeech test-clean, a standard set of clearly read English with official transcripts, and counted a 1.94% word-error rate — 98.1% of words right. Read speech is easier than most real uploads, so treat it as a best case; the method is in our accuracy breakdown. We have not run the same test on the other ten, so this page compares limits and features — not transcript quality.
Only facts from our own code: no account at any point, 10 transcriptions a day for free, files up to 60 minutes and 1.5 GB, no watermark, TXT plus SRT and VTT from every upload, speaker labels for two to eight voices, and about 100 languages detected automatically. Uploads are deleted right after processing and results after 45 minutes (24 hours with a pass). A €5 pass is a one-off payment for 30 days, not a subscription.
Scanning paper, not transcribing audio? Our sister roundup, best free document scanner tools, compares six free options the same honest way.
Yes. You get 10 files a day, any time you like; the per-file ceiling is 60 minutes and 1.5 GB. There is no account, no watermark and no card, and the limit resets daily at midnight rather than being a monthly pool that runs out mid-month. Free transcriptions here are full free automatic transcription, not a preview: the AI transcription free tier is the same model as the paid one, so you can transcribe for free with no catch beyond the daily limit. Ads are what keep the GPU running.
It is our page and our tool is on it, which is why every limit in the table links to the other tool's own pricing page, checked on 2 October 2026, and why there's a section on where the others are better. The best transcription tools here are ranked by their published free-tier limits, not by marketing claims and not by an accuracy test. If you find a limit that has changed, mail hello@shinobitools.com and we will correct it.
There isn't one best transcription software for everyone, so the honest answer to "what is the best free transcription software" is: pick by the limit that bites you, since some tools win on a single feature like local processing or live captions. For hosted tools, ScribeGrab allows the most files a day (10) and Otter.ai the most minutes a month (300). Whisper and MacWhisper have no cap because they run on your own computer. That covers the best free transcription apps, the best free transcription website and the best free transcription services alike: the best transcription apps for you depend on your limit.
Whisper, run locally or through a browser demo like Whisper Web, or MacWhisper on a Mac, because the audio never leaves your device at all. ScribeGrab's upload is deleted right after processing and wiped within 45 minutes (24 hours with any pass, free or paid), but it does reach our server first.
Only the ones you run yourself. Whisper and MacWhisper are free transcribing software you install, the transcription programs free for good and transcription software free of any meter, real transcribe freeware with no time limit, because nobody meters your own computer. Everything running on someone else's server, ScribeGrab included, has a ceiling somewhere. Happy Scribe and Sonix are one-time trials; ScribeGrab stays free with a daily limit instead of a trial, with no expiry.
Three: ScribeGrab, Whisper and MacWhisper. Every other tool in the table asks for a sign-up before you can use the free tier, so among hosted transcription applications free of any sign-up, ScribeGrab is the one. Free video transcription with no membership works the same way: ScribeGrab takes video files without an account.
Two do on their free tier, according to their pricing pages: Descript (720p, watermarked video) and Happy Scribe (watermarked MP4 exports) — a pattern common among free video editors, which usually only drop the watermark on a paid plan. ScribeGrab puts no watermark on anything.
Among the tools with a monthly cap, Otter.ai's 300 minutes a month is the widest, though its limit of 3 file imports for the lifetime of the account makes it a poor fit if you're not joining live calls. ScribeGrab has no monthly pool: 10 files a day, 60 minutes each.
To create a transcript from a video, upload the video file to any tool in the table; you don't need to convert video to audio first. The tool pulls the audio track out and transcribes it, so going from video to transcript is one step, and the same goes for a transcript from audio. To get transcript from video files (or get transcript of video recordings from a phone), or to get video transcript files with timestamps, pick SRT or VTT: those make video transcript files you can load as subtitles. Video transcription free of charge always stops at each tool's own limit; ScribeGrab treats video like audio, while Otter.ai, Notta, Descript and Rev count video minutes against their monthly allowance.
On clear speech, modern speech-recognition models sit around 2% word errors; accents, crosstalk and background noise push that up fast. We measured our own model at a 1.94% word-error rate on 60 fragments of clean, read English (LibriSpeech), about 98.1% of words right. We haven't run the same test on the other tools, so this page doesn't rank the best transcription programs on accuracy; it compares limits and features.
No, it is fully automatic transcription. ScribeGrab is an AI powered transcription tool: an audio AI model does the listening and writing on our own GPU, and nobody reads or edits the text before it's transcribed and sent back to you. Typing the same hour out by hand takes several times as long, which is why people use AI for transcription in the first place. Rev is the one in the table that also sells human transcripts.
Yes. Several tools in the table, ScribeGrab included, transcribe a downloaded call recording the same way as any other audio file. For live call transcription while the call happens, Otter.ai or Notta are built for that instead.
One with long files and no monthly pool. Among hosted tools that is ScribeGrab, with 60 minutes per file and 10 files a day; with no limit at all, Whisper or MacWhisper on your own computer. For a pile of MP3s, the best audio transcription software free of any install is a browser tool, and the best audio transcriber free of any limit is MacWhisper on a Mac or Whisper anywhere.
Whisper and MacWhisper, if you can install software: no cap at all. Of the hosted transcribing tools free to use every day, ScribeGrab allows 10 files and TurboScribe 3. The transcription services free to use in a browser are ScribeGrab, Otter.ai, Notta, Descript, Rev, Happy Scribe, Sonix and TurboScribe; Deepgram is an API for developers.
The hosted ones don't: ScribeGrab, Otter.ai, Notta, Rev, Happy Scribe, Sonix and TurboScribe all work in a browser, so free online transcription is an upload away. Whisper and MacWhisper need installing as transcribing software, and Descript comes as a desktop app as well as in the browser. Among the hosted transcription websites free of charge, ScribeGrab is the one with the widest daily cap, and so the one we built for repeat use; most of the other transcription websites meter minutes per month or offer a one-time trial.
The transcript as TXT and the SRT and VTT subtitles are all free. The paid pass adds longer files (up to 4 hours), results kept for 24 hours, Word export, meeting notes and speaker names. It doesn't change the speech model: free and paid use the same one.
For hosted AI audio transcription free of charge, yes: 60 minutes per file and 10 files a day is the widest free ceiling in the table. Whisper and MacWhisper have no cap at all, but they run on your own computer. Which is the best free AI audio transcription tool, or the best AI transcription for you, still depends on what you need: among free audio transcription tools, each one is a free audio transcription tool with a different catch.
Plain TXT, with SRT and VTT subtitle files made from the same upload. Going in, MP3 transcription free of charge works the same as WAV, M4A and most other audio formats, and video files work too. There's a 1.5 GB and 60-minute limit on audio file transcription.
Yes. ScribeGrab is an online transcription tool, so it runs in a phone browser the same way as on a laptop, with no separate app. A voice memo works like any other audio file, and a full 60-minute free recording fits in one upload. It's a transcription application free to use every day, so you don't need a transcription app free trial or an install. It needs no login, so nothing stays on a shared computer except the files you choose to download.
Usually a few minutes. Getting an MP3 to transcription free of charge takes as long as the file and the queue need; a paid pass only moves you up the queue. It isn't instant: each file needs GPU time.
Your upload is deleted right after processing, and the results are wiped within 45 minutes (24 hours with any pass, free or paid).
Roughly 100. Like any AI voice transcription, accuracy still varies with accent, background noise and crosstalk.
ShinobiTools · KvK 42147182 · info@shinobitools.com