Can you make money transcribing audio?

Short answer: yes, but modestly — and only if you understand the one thing that decides everything about this work. Transcription doesn't pay by the hour. It pays by the audio minute. That single distinction is the whole story of why some people earn a decent side income and most quit within a month.

The pitch is seductive: listen to audio, type what you hear, get paid, work from home. No boss, no commute, no experience required. And it's not exactly false. Thousands of people do earn real money transcribing. But the gap between "earn money" and "earn a living" is wider here than in almost any other online work, and it comes down to math that nobody explains upfront.

Here is the honest version.

The pay structure that decides everything

A transcription platform doesn't pay you for your time. It pays you for the audio's time. A 60-minute podcast episode pays the same fee whether it takes you 45 minutes or four hours to transcribe. That is the entire economics of the job in one sentence.

The standard unit is the audio minute. As of 2026, the going rates look roughly like this: general transcription pays around $0.24 to $0.65 per audio minute depending on the file's difficulty and your grade on the platform. Captioning — typing text synced to video timestamps — pays roughly $0.54 to $0.70 per video minute. Foreign-language subtitling pays more, $0.70 to over $1.00 per minute, but only if you're fluent and pass a separate test.

Now do the beginner's math, because this is the part that matters. A beginner typically takes four to five minutes of their own time to transcribe one minute of audio — pausing, rewinding, looking up spellings, fixing formatting. At $0.40 per audio minute, one audio minute earns you forty cents for five minutes of work. That is under five dollars an hour. The same file, done by someone fast, takes barely over a minute of work per audio minute — and suddenly the effective rate is twenty dollars an hour. Same pay scale. Completely different job.

Your typing speed isn't a skill here. It's your salary.

What the platforms actually pay

The big names are worth knowing, because they set the market.

Rev is the largest and the most discussed. Its advertised range is $0.30 to $1.10 per audio minute, but most general transcription work lands in the lower half of that — $0.30 to $0.65. Rev has also cut rates more than once over the years; longtime freelancers still talk about a reduction years ago that dropped many people's per-minute pay by around 40 percent overnight, with no negotiation. That history tells you something important about the relationship: you're a contractor with no pricing power, on a rate that changes when the platform decides it changes.

GoTranscript pays around $0.40 to $0.65 per audio minute and accepts transcribers from most countries, which makes it one of the more accessible options worldwide. TranscribeMe pays roughly $15 to $22 per audio hour for experienced workers, with higher rates for specialist work. Scribie pays about $5 to $25 per audio hour depending on file difficulty and your accuracy score. Way With Words and GMR Transcription pay toward the higher end per minute; Daily Transcription offers a higher minimum than most.

None of these will make you rich, and the honest ones don't pretend otherwise. A realistic beginner working three to four hours a day, five days a week, earns somewhere around £400 to £700 a month once up to speed — that's the number nobody selling a transcription "course" wants to tell you. Data from freelancer communities suggests the average Rev worker earns around $156 total after their first 15 to 20 jobs, while the highest-paid seasoned workers clear a little over $1,000 a month. Those are the real brackets: a few hundred a month for most, a thousand-plus for the fast and persistent.

Speed is the only raise you'll ever get

Nobody is going to give you a raise in transcription. The rates are the rates. The only variable you control is how fast you work — which means speed-building is the entire career ladder.

The floor is typing: 80 words per minute is roughly where this work starts feeling viable, and the people earning real money type faster. Beyond raw speed, the tools matter enormously. A foot pedal for play-pause-rewind — so your hands never leave the keyboard — is the single highest-return purchase in this job, and it costs less than dinner. Text expanders turn your frequent phrases into keystrokes. Good headphones turn mumbling into words. Transcription software with adjustable slowdown without pitch distortion saves minutes per file.

Then there's the unglamorous stuff: learning style guides cold, building a personal dictionary of terms in the niches you work, and — the one beginners resist — cherry-picking files. Clear audio with one speaker pays the same per minute as a noisy panel discussion with three accents, but takes a third of the time. Experienced transcribers grab the clean files fast and leave the terrible ones for whoever is desperate. That isn't gaming the system. That is the system.

Specialization is the other raise. Medical and legal transcription pay meaningfully more — $15 to $40 an hour in some niches — but require training, terminology, and often certification. It's a real path upward. It's just a different job than "type what you hear," and it takes months to qualify for.

Getting in: tests, grades, and the first month

Every platform makes you audition, and the audition is harder than the job. Rev's entry test — a quiz plus a sample transcription graded against their style guide — filters out most applicants on the first try. The ones who pass usually failed once first. This isn't a bug; it's the business model. Platforms need to know you can follow formatting rules before they trust you with client files.

Then comes the grade system, which is the part nobody explains. Most platforms rank freelancers — Rev uses letter-style quality grades, others use scores or tiers — and your grade decides which files you can claim. Beginners get the dregs: the noisy, the accented, the difficult. Do good work and your grade rises, unlocking cleaner files that take less time for the same per-minute pay. In other words, the job literally gets easier and better-paid the longer you survive it. The catch is surviving the first month, when you're slow, the files are bad, and the effective hourly rate is embarrassing.

Go in expecting month one to be training, not income. If you can stomach that — and keep your accuracy high while you're slow — month three looks completely different from month one. Most people quit in week two. That's not a tragedy; it's just the filter working as designed.

The AI problem nobody can dodge

Here's the honest threat: automated transcription keeps getting better, and it's eating this market from the bottom up.

AI transcription now hits around 90 percent accuracy in ideal conditions — clear audio, one speaker, no jargon. That's not good enough for legal or medical work, but it's good enough for a huge chunk of the general transcription market, especially since clients can pay a human a fraction of the old rate just to clean up the AI's draft. That cleanup work pays less than transcription from scratch, and it's growing as a share of available jobs.

This doesn't mean human transcription disappears. It means the middle hollows out. The work that survives is the work AI does badly: terrible audio, heavy accents, overlapping speakers, specialist terminology, strict accuracy requirements, and anything where errors have consequences. The premium moves toward difficulty and trust, away from volume.

The practical takeaway: if you're entering transcription now, don't build your plan on easy general files. Those are exactly the files AI does cheapest. Build toward the hard stuff — the files people complain about — because that's where the human premium will live.

Who this is actually good for

Transcription is a genuinely good fit for a specific kind of person, and a trap for everyone else.

It's good for you if you type fast already, you like solitary detail work, you need fully flexible hours — the school-run parent, the night owl, the person in a country where $500 a month goes far — and you're willing to treat the first two months as paid training at bad wages while your speed climbs. For that person, it's honest work with honest, if modest, pay.

It's a trap if you need reliable money from week one, because the per-minute structure punishes beginners hardest exactly when they're slowest. It's a trap if you hate repetitive detail work, because this job is nothing but repetitive detail work. And it's a trap if you're looking at it as a career rather than a side income, because the ceiling is low and AI is lowering it.

The kindest thing anyone can tell you about transcription is also the truest: it's a fine way to earn a few hundred dollars a month from your couch, a hard way to earn a thousand, and nobody's path to wealth. Go in with that expectation and the right typing speed, and it pays. Go in expecting easy money, and the pay structure will teach you the lesson itself — one audio minute at a time.