TL;DR

Meeting transcription and file transcription are different jobs needing different tools. A meeting bot that joins Zoom calls is useless for transcribing a recorded interview. A file transcription service is useless for live meeting notes. Identify which job you need before evaluating anything.

• Fathom is the best free meeting transcription tool most people have never heard of. Free for individuals with no minute limits. Joins Zoom, Google Meet, and Teams. Generates AI summaries, action items, and searchable transcripts. No watermark. The free individual plan is a complete product, not a capped trial.

• Otter.ai is the most recognized meeting transcription tool. Free for 600 minutes per month. OtterPilot joins calls automatically and produces AI summaries organized by topic. Good collaboration features on paid plans. The 600-minute limit fills up faster than expected for heavy meeting schedules.

• Fireflies.ai is the right choice for sales teams. Meeting transcription with direct CRM sync to Salesforce, HubSpot, Pipedrive, and others. Searchable meeting history across all recorded calls. Free plan is 800 minutes total storage (not per month), so it runs out and you need to upgrade or lose access to older transcripts.

• Whisper is free, open source, and often more accurate than paid services for audio files. Trained on 680,000 hours of diverse audio. Handles accents, technical vocabulary, and long files better than most commercial services. Requires Python and command-line setup. MIT license allows commercial use.

• MacWhisper gives you Whisper accuracy on Mac without any technical setup. Drag and drop an audio or video file, pick a model, click transcribe. Free basic version. Pro at $19.99 one-time (no subscription) adds large high-accuracy models, batch processing, and translation. Everything runs locally on your device.

• Rev is for when accuracy cannot be wrong. Automated transcription at $0.25/min is competitive. Human transcription at $1.50/min adds real transcriptionists who review and correct the output, producing 99 percent accuracy even on difficult audio with background noise, heavy accents, and multiple speakers. Pay-per-use with no subscription.

• Descript is for editing audio and video by editing text. Transcribe a recording, then delete words from the transcript to cut the corresponding audio. Remove all filler words in one click. Fix mispronounced words using AI-generated voice. Completely different workflow from any other tool. Free 1 hour per month. Hobbyist at $24/month.

• Sonix is the pay-as-you-go choice for researchers and journalists. $22 per hour with no monthly fee. Good accuracy, 38 languages, inline correction, speaker labels. Popular in academic settings where transcription volume varies across projects.

• Privacy matters for sensitive recordings. Cloud services (Otter, Fireflies, Sonix, Rev) upload your audio to their servers. Local tools (Whisper, MacWhisper, Superwhisper) keep everything on your device. For medical, legal, or confidential recordings, use local tools. Most cloud transcription services are not HIPAA-compliant by default.

• Speaker identification accuracy varies significantly. All automated tools make mistakes when voices are similar. If who said what is critical, budget time for manual correction or use Rev's human transcription.

• Always test with your actual audio before committing. Benchmark accuracy claims come from clean audio. Your specific conditions (background noise, accents, technical vocabulary) may produce very different results. A 5-minute test clip saves you from discovering problems after paying for hours of transcription.

——————————————————————————————————————————-

Best Transcription Apps in 2026: What Actually Works for Meetings, Files, and Everything Else + Reddit Insights

Last updated: July 2026

The transcription market split into two very different products a few years ago and most roundups still treat them as the same thing.

One product transcribes your meetings in real time: joins the call, takes notes, sends a summary when it ends. The other product transcribes audio and video files you already have: interviews you recorded, podcasts you need indexed, videos you need captioned.

These are different jobs. The best tool for one is often the wrong tool for the other. A meeting tool that joins Zoom calls is useless for transcribing a recorded interview. A file transcription service is useless if you need live notes in your next meeting in twenty minutes.

This guide covers both categories, names the best tool in each, and covers the one free option that Reddit keeps recommending that most roundups do not feature prominently: Fathom, which transcribes meetings free with no minute limits, and Whisper, which transcribes audio files free with accuracy that beats most paid services.

Which type of transcription do you need?

Before picking any tool, identify which job you are trying to do.

Real-time meeting transcription: You need a bot that joins your Zoom, Google Meet, or Teams call, transcribes while the meeting happens, and delivers notes and a summary when it ends. Tools: Otter.ai, Fireflies.ai, Fathom, Grain.

Audio or video file transcription: You have a recording (MP3, MP4, WAV, M4A) and you need it converted to text. Tools: Whisper, MacWhisper, Sonix, Rev, Happy Scribe, Notta.

Real-time dictation: You want to speak and have text appear immediately wherever your cursor is, like a dictation feature. Tools: Superwhisper (Mac), Windows Voice Access, Apple Dictation. (Covered in our dictation software guide.)

Captions for video content: You are publishing a video and need closed captions or subtitles. Tools: Descript, Rev, Happy Scribe, YouTube's automatic captions.

Most people need one or two of these, not all four. Read the section that matches your use case rather than evaluating every tool.

Quick-pick table

Fathom: the free meeting transcription nobody knows about

Fathom is the tool Reddit keeps recommending that does not appear in most transcription roundups, probably because it does not charge anything and therefore does not run affiliate programs or pay for editorial placement.

Fathom records and transcribes Zoom, Google Meet, and Microsoft Teams calls. It generates AI summaries organized by topic. It identifies action items. It clips specific moments from the recording. It integrates with HubSpot, Salesforce, and Notion. The transcripts are searchable. You can share clips with people who were not on the call.

And the individual plan is completely free with no minute limits.

Not "free with 600 minutes per month." Not "free for your first 10 calls." Free. No limits. For individuals.

The team plan at $19 per user per month adds collaborative features, more CRM integrations, and team analytics. But for a solo professional or anyone who just wants their own meetings transcribed and summarized, the free plan is a complete product.

The catch: Fathom works by joining the call as a participant named "Fathom Notetaker." Some meeting participants find this unexpected or uncomfortable. Fathom's notetaker bot is visible to everyone in the meeting, which is appropriate for consent reasons but requires a brief explanation to meeting participants who have not seen it before.

Also: Fathom currently supports Zoom, Google Meet, and Teams only. If you use a different video conferencing tool, Fathom does not work.

Best for: Anyone who wants free unlimited meeting transcription and AI summaries. Individual professionals who want meeting notes without paying.

Pricing: Free for individuals. Team at $19/user/month.

Otter.ai: the most recognized meeting transcription tool

Otter.ai has been around since 2016 and built the category of real-time AI meeting transcription. It is the name people think of when they think of this tool type, and for most use cases it is genuinely excellent.

The free plan gives you 600 minutes of transcription per month. That is 10 hours, which covers roughly two hours of meetings per week. For many individual users, the free plan is enough.

The AI features have matured significantly in 2025 and 2026. OtterPilot joins your Zoom, Google Meet, or Teams call automatically, transcribes in real time, and when the meeting ends produces a summary organized by key topics, a list of action items with names attached, and a searchable full transcript. The summary quality is genuinely useful: it surfaces what was decided and who agreed to do what, not just a list of everything that was said.

The collaboration features on the Business plan let team members access and annotate shared transcripts. You can highlight a section, add a comment, and tag a colleague. For teams that review meeting notes together, this works well.

Where Otter has limitations: the speaker identification (who said what) is less reliable when multiple people have similar voices or when audio quality varies. Background noise degrades accuracy noticeably. And the 600-minute free limit fills up faster than people expect when they use it for every meeting.

For most individuals who attend moderate numbers of meetings, the free tier is the right starting point. If you hit the limit, evaluate whether $16.99 per month for unlimited is worth it before switching tools.

Best for: Knowledge workers who want real-time meeting transcription with AI summaries. Individuals starting with meeting AI who want to test before paying.

Pricing: Free for 600 min/month. Pro at $16.99/month. Business at $30/user/month.

Fireflies.ai: for sales teams that need CRM integration

Fireflies is a meeting transcription tool built with sales teams specifically in mind. The CRM integration is the differentiator.

After a call transcribed by Fireflies, you can push the transcript, summary, and key moments directly to your CRM. Salesforce, HubSpot, Pipedrive, Zoho, and others. The call activity, the summary, and the action items appear on the contact or opportunity record automatically. For sales operations where keeping the CRM current after every call is a requirement, this removes a category of manual work.

The searchable meeting history is another Fireflies strength. You can search across every meeting the tool has ever recorded for a specific topic, a competitor name, a product feature, or a person. "When did we discuss changing the pricing?" becomes a search rather than a memory exercise.

The free plan includes 800 minutes of storage, which is more than Otter's 600-minute monthly limit. The difference is that Fireflies' free minutes represent storage, not a monthly allowance: 800 minutes total, not per month. Once you use those 800 minutes, you need to upgrade or lose access to older transcripts.

Pro at $18 per month adds unlimited storage, unlimited AI summaries, more CRM integrations, and the smart search feature. Business at $29 per month adds conversation intelligence, speaker talk-time analytics, and coaching tools.

Best for: Sales teams that want meeting transcription integrated with their CRM. Revenue operations professionals who need searchable meeting history. Any team where post-call CRM updates are a regular requirement.

Pricing: Free (800 min storage). Pro $18/month. Business $29/month.

Whisper: the most accurate free transcription for audio files

OpenAI released Whisper as an open-source model in September 2022. It has since become the reference accuracy standard for automated transcription, and most of the best transcription tools on this list use it or a derivative of it under the hood.

For transcribing audio files, Whisper running locally on your own computer often produces more accurate results than paid commercial services. It handles accents, technical vocabulary, background noise, and multiple languages better than most alternatives. And it is free, with an MIT license that allows commercial use.

The trade-off is setup. Running Whisper directly requires Python and a command-line environment. The process: install Python, install the Whisper library via pip, download a model (different sizes trade off accuracy for speed), and run it against your audio file. For a developer or someone comfortable with terminal commands, this takes about 30 minutes the first time. For everyone else, it is a barrier.

The accuracy advantage is most pronounced in three situations:

First, unusual vocabulary. Whisper was trained on 680,000 hours of diverse audio and handles domain-specific terms (medical terminology, legal language, technical jargon) better than tools trained on more general data.

Second, accents and non-standard speech patterns. Whisper's training data was intentionally diverse, giving it stronger handling of accents than most commercial services that were primarily trained on American and British English.

Third, long files. Commercial transcription services sometimes drift in accuracy over long recordings. Whisper maintains consistency across files of any length.

For anyone comfortable with the setup, Whisper is worth the effort. For everyone else, MacWhisper or Superwhisper provides the same accuracy through a clean interface without touching the command line.

Best for: Developers and technical users comfortable with Python. Anyone who needs the highest possible accuracy for audio file transcription at zero cost. Users with privacy requirements who cannot send audio to cloud servers.

Pricing: Free and open source.

MacWhisper: Whisper accuracy without the technical setup

MacWhisper is a Mac app that wraps OpenAI's Whisper model in a clean interface. Drag an audio or video file onto the app, select a model size, click transcribe, and get text output with timestamps.

The free version handles basic transcription with smaller models. The Pro version at $19.99 is a one-time purchase (no subscription) and adds the large high-accuracy Whisper models, batch processing for multiple files, export in multiple formats, and translation from 90+ languages to English.

The batch processing feature is genuinely useful for researchers, journalists, and podcast editors who regularly transcribe multiple files. Select 20 interview recordings, set them to process overnight, and wake up to completed transcripts.

Everything runs locally on your Mac. No audio leaves your device. No account required. No API keys. This is the privacy case: your recordings, your device, your data.

Performance scales with your Mac hardware. On an M-series Apple Silicon Mac, even the large high-accuracy model processes audio significantly faster than real time. On older Intel Macs, the large model can be slower than real time and the medium model is often the practical choice.

Best for: Mac users who want Whisper accuracy without command-line setup. Researchers, journalists, and content creators who transcribe multiple files regularly. Anyone who needs local processing for privacy.

Pricing: Free basic version. Pro at $19.99 one-time.

Rev: when accuracy cannot be wrong

Rev is the choice when you need the highest possible accuracy and you cannot tolerate errors because the transcript is going to court, into a medical record, or into published journalism.

The automated transcription at $0.25 per minute is competitive with other commercial services. The human transcription at $1.50 per minute is what sets Rev apart. Real transcriptionists review and correct the AI output, producing accuracy levels that no automated service consistently matches for difficult audio.

Human transcription from Rev produces results that are typically 99 percent accurate even for poor-quality audio: background noise, multiple simultaneous speakers, heavy accents, heavily technical vocabulary. This level of accuracy requires human review because no AI model yet handles all of these challenges consistently.

The turnaround for automated transcription is minutes. For human transcription, Rev's standard service typically delivers within 24 to 48 hours, with rush delivery options available.

The pay-per-use model with no subscription means you only pay for what you use. For infrequent transcription of high-stakes content, this is more economical than a monthly subscription you might not fully utilize.

Caption generation at $1.50 per minute handles SRT and VTT files for video captions, with human accuracy on the timing and text.

Best for: Legal proceedings, medical documentation, journalism, and any context where a transcription error creates real risk. Users who regularly have poor-quality audio or difficult accents that automated services handle poorly.

Pricing: Automated $0.25/min. Human $1.50/min. Captions $1.50/min.

Descript: when you want to edit audio and video by editing text

Descript is in a different category from everything else on this list. It is not just a transcription tool. It is an audio and video editor built around the idea that the transcript is the edit.

You import a video or audio file. Descript transcribes it. You then edit the recording by editing the transcript. Delete a word in the transcript and the corresponding audio is removed from the recording. Select a section of transcript and cut it out and the video cuts seamlessly. Add a pause by clicking between words in the transcript.

This workflow is genuinely transformative for podcast editing, video content creation, and anyone who edits their own audio. The traditional approach requires scrubbing through a timeline looking for the sections you want to cut. Descript's approach lets you read the transcript and make decisions like editing a document.

AI-powered cleanup removes filler words (um, uh, like, you know) from the recording in one click. Overdub generates audio in your own voice from text so you can fix a mispronounced word without re-recording. Studio Sound cleans up audio quality from noisy environments.

The transcription quality is excellent and powers all of the editing features. An error in the transcript creates an error in the edit, so Descript invests heavily in transcription accuracy.

The free plan allows 1 hour of transcription per month. The Hobbyist plan at $24 per month allows 10 hours. For a weekly podcast, 10 hours per month covers roughly five 2-hour interview recordings.

Best for: Podcast producers, video creators, and content editors who want to edit audio and video by editing text. Anyone who spends significant time editing recordings and wants a faster workflow.

Pricing: Free 1hr/month. Hobbyist $24/month. Creator $40/month.

Sonix: for researchers and journalists who pay per use

Sonix is the pay-as-you-go choice most recommended in Reddit threads from researchers, qualitative analysts, and journalists who transcribe interviews.

The pricing is $22 per hour of audio (roughly $0.37 per minute) with no subscription required. You buy credits and spend them as needed. There is no monthly fee eating your budget when you are between projects.

The accuracy is strong and the interface is clean. You can correct the transcript inline, add speaker labels, and export in multiple formats. The speaker identification separates who said what reasonably well for clearly distinct voices.

The 38-language support and the reasonable pay-per-use economics make it popular in academic settings where researchers might transcribe interviews in multiple languages across a project without a predictable monthly volume.

For teams that need collaboration on transcripts, Sonix offers subscription plans starting at $5 per month plus $5 per hour of audio. These plans add sharing, team workspaces, and workflow tools.

Best for: Researchers, qualitative analysts, and journalists who transcribe interviews irregularly. Academic projects. Anyone who wants pay-per-use pricing without a monthly commitment.

Pricing: $22/hour pay-as-you-go. Subscription from $5/month plus $5/hour.

The accuracy question: which tool is actually most accurate?

Every transcription service publishes accuracy claims. Most quote word error rates from benchmarks that are not representative of real-world conditions.

The honest answer from actual user testing is more nuanced.

For clear English audio with a single speaker in a quiet environment, all major automated transcription services (Whisper, Otter, Fireflies, Sonix, Rev automated) achieve accuracy in the 95 to 99 percent range. The differences are small.

For challenging audio (background noise, multiple simultaneous speakers, heavy accents, technical vocabulary, poor microphone quality), the differences become significant. In challenging conditions, Whisper-based tools often outperform commercial services because of their training data diversity. Rev's human transcription outperforms all automated options.

Speaker identification quality varies even more. Correctly attributing speech to the right speaker when voices are similar is a hard problem. All automated tools make mistakes here. If who said what is critical to your use case, budget extra time for manual correction or use Rev's human transcription.

The practical advice: for your first time with any service, test it with a sample of audio representative of what you will actually transcribe. A 5-minute test clip saves you from discovering accuracy problems after paying for 10 hours of transcription.

Privacy: who can read your transcripts?

If your recordings contain sensitive information (patient conversations, legal discussions, confidential business strategy, personal matters), where the audio goes matters.

Cloud services (Otter, Fireflies, Sonix, Rev): your audio is uploaded to their servers for processing. Review each provider's data handling policies and data retention practices before using them for sensitive recordings.

Local processing (Whisper, MacWhisper, Superwhisper): audio never leaves your device. Nothing is uploaded. No account needed in some tools. This is the correct choice for anything confidential.

HIPAA compliance for medical recordings requires specific agreements with transcription vendors. Most general transcription services are not HIPAA-compliant by default. If you transcribe patient conversations, verify HIPAA BAA availability with any service you consider.

Common questions

What is the most accurate transcription app? For automated file transcription, Whisper-based tools (Whisper locally, MacWhisper, Superwhisper) offer the best accuracy especially for accents and technical vocabulary. For highest accuracy regardless of audio quality, Rev's human transcription service is the most reliable.

Is there a genuinely free meeting transcription tool? Yes. Fathom is free for individuals with no minute limits. Otter.ai's free tier covers 600 minutes per month.

Can I transcribe a video file? Yes. Most transcription services accept MP4, MOV, and other video formats in addition to audio files. MacWhisper, Sonix, Happy Scribe, and Descript all handle video files directly.

How accurate is automated transcription? For clear audio with a single speaker, 95 to 99 percent word accuracy is typical for modern services. For challenging audio with background noise or heavy accents, accuracy can drop to 80 to 90 percent. Test your specific use case before committing.

What languages do transcription tools support? Whisper supports 99 languages with good quality on major languages. Most commercial services support 30 to 60 languages. Happy Scribe is particularly strong for European languages. Check your specific language before choosing.

Can transcription software identify different speakers? Most tools offer speaker diarization (labeling who said what). Accuracy varies significantly across tools and is lower when speakers have similar voices. For perfect speaker attribution, budget time for manual review or correction.

Is it legal to record and transcribe meetings? This depends on jurisdiction. In the US, rules vary by state. California, for example, requires all-party consent. For meetings with people in multiple states or countries, the strictest applicable law applies. Always inform participants that a meeting is being recorded.

The bottom line

For meeting transcription: start with Fathom free (unlimited, no minute caps) if you use Zoom, Google Meet, or Teams. Use Otter.ai if you want more polished AI summaries or need features beyond Fathom's free tier.

For audio file transcription: use Whisper locally (or MacWhisper on Mac) for the best accuracy at no cost. Use Sonix for pay-as-you-go file transcription with a clean interface. Use Rev when accuracy must be near-perfect and you can pay for human review.

For podcast and video editing: Descript changes the workflow entirely. Worth trying before buying.

And for anything sensitive: keep audio on your own device with MacWhisper or Whisper rather than uploading to cloud services.

Transcription accuracy and pricing in this category change frequently. Verify current details at each provider's website before subscribing.