Sonix: reviews and analysis

TLDR: Sonix turns audio and video into text in more than fifty languages, with word-level timestamps, speaker separation and an editor where you fix the text while the recording plays. It is the best-connected tool in this segment: it reaches into Zoom, Teams, Drive, Dropbox and Adobe Premiere without anyone writing code. Its weak spot is not what it produces but what it charges: the structure mixes a monthly fee with loose hours, translation is billed on top, and the site contradicts itself about the hourly rate.

What Sonix is and how it works

Sonix is a browser-based automated transcription service that has been running since 2017. You upload an audio or video file, pick the language and get back a document in which every word is anchored to a point in the recording. That anchor is what makes the rest work: click a word in the text and the audio jumps there.

The useful comparison is not against a bare speech recognition model but against everything wrapped around it. A model hands back a string. Sonix hands back a document with speakers split apart, low-confidence passages colour-coded so you know where to look first, a custom dictionary where you load the names and jargon of your field, and a long list of output formats covering SRT and VTT subtitles plus the files Premiere, Final Cut and DaVinci Resolve expect.

Spanish is covered in detail, and that matters more than it sounds. The tool separates Peninsular Spanish from Latin American Spanish and recognises regional varieties including Mexican, Caribbean, Andean and Rioplatense. Public reviews carry the same surprise more than once from people transcribing Argentinian interviews who expected the machine to choke.

On top of transcription sits a translation layer covering more than fifty languages and an API with webhooks for anyone who would rather wire up their own circuit. That API is real, with public documentation and operations to upload media, request the transcript in several formats and order translations, so Sonix works as an application and as plumbing.

What using it is like day to day

The normal flow is short. Upload, wait a few minutes, open the editor. What decides whether the next hour is pleasant or grim is not the tool but the source audio: on clean single-speaker recordings the text comes out nearly finished, and on noisy meetings with people talking over each other you will be doing passes.

That distinction is not a vendor excuse, third-party readings back it. A comparison with its own method, published by a science journalism outlet, sent the same two interviews to several services and printed the raw output side by side, Sonix’s errors and its missing-sentence gaps included. A signed review in the technology press gave it 4.5 out of 5 and complained that it over-corrects on grammatical grounds, which is a polite way of saying it sometimes tidies up what you actually said.

The editor is built for shared work. You can leave comments, assign sections to somebody else and organise files into folders, which makes it reasonable for a newsroom or a qualitative research team. The connectors it ships with are the other reason people stay: if your recordings already live in Drive, Dropbox or Zoom, the file walks in by itself.

Two absences are worth holding in mind. There is no mobile app, so recording on a phone and transcribing in Sonix are two separate steps you have to join yourself. And automatic speaker recognition still turns up labelled as a beta feature in third-party comparisons, which fits the reviews complaining that roundtables need labels reassigned by hand.

Pricing and plans

A warning belongs before the figure. Sonix sells two ways at once, and that is what tangles any comparison.

On one side there is a pay-per-use plan with no subscription, at 10 dollars per hour of uploaded audio, prorated to the second, which is the cheap route for anyone transcribing one interview a month. On the other side there are subscriptions, and the entry one is the Core plan from 25 dollars a month with five hours included. Hours beyond your allowance are billed again at 10 dollars. Our structured data records those 25 dollars because it is the monthly fee comparable with the rest of the segment, but if you transcribe rarely, the figure that touches you is the other one.

Then comes the part that annoys people who already pay. Translating is not included in transcribing: it is billed separately and at the same rate, so a file you need in two languages costs twice. Several reviews describe the surprise with the same sentence, the one about not knowing what they would pay until they had paid it.

The inconsistency reaches the site itself. The structured data on the pricing page declares a plan with a different name and a different amount from the one the table shows, and the homepage advertises an hourly rate different from the one that table publishes. Neither of those is the good figure. The good figure is the one in the plan table, and that is what we record.

There is no permanent free plan. What there is instead is a thirty-minute trial of transcription with no card and no expiry date, which runs out when you spend it.

Who it is for (and who it is not for)

Sonix pays off when you transcribe in several languages every month and the text has to land somewhere specific: a video edit, a research archive, an automated pipeline. That combination of languages, output formats and connectors is what no single rival gives you all at once.

It is not for someone who transcribes occasionally and wants to know the bill in advance. There, predictable pricing wins, and Transkriptor starts under ten dollars a month with an allowance written into the table. Nor is it for anyone handling sensitive material who needs European invoicing and infrastructure, which is Amberscript territory. And if subtitles with a human safety net are the job, look at Happy Scribe first. The rest of the ground is in the best AI audio tools.

Alternatives to Sonix

The split across this segment is fairly clean. Gladia is the option for anyone embedding transcription inside their own product with no interface wanted. Transkriptor wins on entry price and on surfaces, with mobile and a browser extension. Amberscript wins on European fit. Happy Scribe wins on subtitles and on human correction bought by the minute. Sonix wins when what weighs is the spread of languages and the number of places the text has to land.

Frequently Asked Questions

What is the real entry price?

It depends how you buy. The cheapest subscription is Core from 25 dollars a month with five hours included. Without a subscription, pay-per-use is 10 dollars per hour of uploaded audio, and that is the cheaper route if you transcribe rarely.

Does it handle Spanish well?

Yes, and with a distinction other tools skip: it separates Spanish from Spain from Latin American Spanish and recognises specific regional varieties. On clean audio the result is solid, and on noisy recordings or overlapping speech you will be editing, exactly as with every other tool in this segment.

Is translation included in the price?

No. Translation is billed separately from transcription and at the same hourly rate, so a file you need in two languages is paid for twice. It is the most concrete pricing complaint in the reviews.

Is there a free plan?

There is no permanent free plan. There is a thirty-minute transcription trial, no card and no expiry, meant for checking quality against your own material before deciding.

Is it any good for subtitling video?

Yes. It exports subtitles as SRT and VTT among other formats, can burn them into the video and delivers files that open in Premiere, Final Cut and DaVinci Resolve. If subtitling is your main task rather than a side one, it is worth comparing against Happy Scribe, which has the most developed subtitle editor of the group.