Sonix
Tool · this deep dive is built from the official pages of the platform itself and of the tools that support it.
Overview
Sonix transcribes audio and video into text automatically, marking speakers and timecodes. Composition from the pricing page and the reference file llms.txt, which the tool publishes for AI crawlers and its robots.txt points to.
The transcript is edited in the browser; from it come subtitles and captions, as SRT and VTT files or burned into video. Declared separately: automatic translation of the transcript, AI analysis and summarisation, search across all transcripts, a REST API and an MCP server.
Languages are where the tool's own numbers disagree, so all get named. The homepage title says "49+ Languages"; llms.txt says 50+ for transcription and 40+ for translation; the pricing page says 54+ and 55+. Four numbers, one site.
A list by name does exist: the languages page gives 54. Arabic, Armenian, Bashkir, Basque, Belarusian, Bengali, Bulgarian, Catalan, Chinese (Cantonese), Chinese (Mandarin), Croatian, Czech, Danish, Dutch, English, Esperanto, Estonian, Finnish, French, German, Greek, Hebrew, Hindi, Hungarian, Indonesian, Italian, Japanese, Korean, Latvian, Lithuanian, Malay, Marathi, Mongolian, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Serbian, Slovak, Slovenian, Spanish, Swahili, Swedish, Tagalog, Tamil, Thai, Turkish, Ukrainian, Urdu, Uyghur, Vietnamese, Welsh.
All six of our catalogue's languages are there: Russian, English, Spanish, Chinese in two rows (Mandarin and Cantonese), Hindi and Arabic. We treat the named list as primary and record the clash of round numbers as it stands. A list of translation languages does not exist: 55+ is named, the names are not.
Which platforms it works with
Two things to separate: where the tool sends files, and what it connects to.
Export formats from the pricing page: DOCX, PDF, TXT, SRT and VTT. The tool's own llms.txt claims "30+ formats", so those five are not the whole list. Subtitles come as SRT and VTT and burned into video.
Integrations: Zoom, Microsoft Teams, Google Meet, Webex, Zapier, Adobe Premiere, plus a REST API and an MCP server.
Direct publishing to social networks does not appear on the pages read. The tool hands out files and hands them into editing suites; posting is yours.
Pricing
From the pricing page, US dollars.
Pay As You Go — 10 USD per hour of audio, no subscription. Core — 25 USD a month or 275 USD a year, with one month free on annual. Advanced — 50 USD a month or 550 a year. Pro — 80 USD a month or 880 a year. Enterprise — on request.
Two surcharges are named separately and apply over any subscription: extra hours at 10 USD each, exactly the Pay As You Go rate, and an extra team seat at 25 USD a month.
The minimum among subscriptions is 25 USD a month on Core. The absolute minimum payment is different: 10 USD for one hour on Pay As You Go, no subscription at all. A subscription is not the only way to pay here.
What is free
There is no permanent free plan; the pricing page shows no free step.
What exists is a 30-minute trial: "30 min free trial", with "No credit card required" and "Cancel anytime".
Those 30 minutes are a one-off, not a monthly quota: given once, not renewed. Once spent, everything after is paid. For scale, it is half of the hour that costs 10 USD on Pay As You Go and a tenth of Core's monthly quota.
Restrictions
Counting runs in hours per month and gigabytes of storage, both tied to the plan.
Pay As You Go — hours bought singly, 5 GB storage, one user. Core — 5 hours a month, 25 GB. Advanced — 20 hours a month, 50 GB. Pro — 40 hours a month, 100 GB. Hours beyond the package cost 10 USD each on any subscription.
Then the rule that doubles the bill for translation, on the pricing page word for word: "automated transcription is an extra charge. It is charged at the same rate as your transcription rate". A translated hour is deducted as another hour, so an hour-long recording transcribed and translated eats two hours of quota; on Core, with its five, that is half the month.
No limit on the length of a single file is named anywhere.
Rights to the output
No direct answer — marked absent. The terms of use do not address rights to uploaded media or to the resulting transcript and translation at all.
What they do contain is a ban on commercially exploiting the service itself: "you agree not to display, distribute, license, perform, publish, reproduce, duplicate, copy, create derivative works from, modify, sell, resell, exploit, transfer or upload for any commercial purposes, any portion of the Service".
The key words are "any portion of the Service": the ban reaches parts of the service, not the text made for you. Nor is there express permission — on commercial use of a transcript the terms are silent both ways. Nothing differs by plan.
No requirement to label a transcript as machine-made: no watermarks on subtitles, no metadata, no C2PA in the terms, the policy or the prices.
Do they train on your data
No — and this is one of the few cases where a tool denies training in plain words: "Sonix does not use your content or Customer Data to train our machine learning or generative AI models".
But where the denial sits matters as much as the denial. The sentence is in the privacy policy, not the terms of use. In the terms, training is not raised at all — the word does not appear there in any form.
The difference is not formal: the terms are a contract, a privacy policy a statement about how data is handled. Anyone looking for the obligation in the contract will not find it. We call the fact confirmed, with the address given exactly: the denial exists, and it lives in the policy.
Separately stated is what happens when a third-party AI application connects over OAuth: Sonix passes "only the data necessary to fulfill the requests you initiate through that application", and from there the other party's policy applies.
There is no opt-out as a separate mechanism, and none is required, since training on customer data is denied by default — hence the partial mark. The only revocation described concerns third-party connections: you may "revoke a connector's access at any time through your Sonix account settings or by contacting us". That is withdrawing an integration's access, not opting out of training.
How you earn with it
On commercial use of a transcript the terms are silent: neither permission nor ban. The one commercial ban concerns parts of the service itself, not the text made with it.
Two things bear on work for hire. Training on customer content is denied, so recordings you transcribe for a client do not, by the tool's own statement, go into models — the denial sitting in the privacy policy. And translation counts as another hour, so transcription-plus-translation jobs cost double quota, or 10 USD per extra hour.
What it will not do
Accept other people's material it will not. The terms forbid uploading anything that "infringes any intellectual property or other proprietary rights" and anything "you do not have a right to upload under any law or under contractual or fiduciary relationships". The ban is framed through rights and legality, not through a list of topics.
And here is what the terms lack, worth saying plainly: no clause requires consent from the people recorded. The rule on third-party rights exists, a direct requirement to notify those recorded does not. For a tool used to transcribe other people's speech that is a visible gap.
Translate free it will not: translation is billed at the transcription rate and takes another hour. Give a permanent free plan it will not: only the one-off 30 minutes. Publish a transcript or subtitles to social networks it cannot — no such export.
Guarantee the training commitment in the contract it will not: the denial is written into the privacy policy, not the terms. Answer for data sent into a third-party AI application over OAuth it will not either — the other party's policy applies from there.
Verified data
The checked data this deep dive rests on.
About the platform
automatic transcription of audio and video into text with speaker labels and timecodes, a transcript editor in the browser, automatic subtitles and captions (SRT/VTT and burnt into the video), automatic translation of the transcript, AI analysis and summarisation, search across transcripts, a REST API and an MCP server
the composition is taken from the pricing page and the service's own llms.txt reference file (https://sonix.ai/llms.txt), opened expressly to AI crawlers in robots.txt
source, checked 2026-08-07
Sources diverge: THERE IS AN EXACT LIST — the /languages page names 54 languages: Arabic, Armenian, Bashkir, Basque, Belarusian, Bengali, Bulgarian, Catalan, Chinese (Cantonese), Chinese (Mandarin), Croatian, Czech, Danish, Dutch, English, Esperanto, Estonian, Finnish, French, German, Greek, Hebrew, Hindi, Hungarian, Indonesian, Italian, Japanese, Korean, Latvian, Lithuanian, Malay, Marathi, Mongolian, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Serbian, Slovak, Slovenian, Spanish, Swahili, Swedish, Tagalog, Tamil, Thai, Turkish, Ukrainian, Urdu, Uyghur, Vietnamese, Welsh. All six languages of the catalogue are there: Russian, English, Spanish, Chinese (Mandarin and Cantonese), Hindi, Arabic. The figures on the site disagree: the home page heading says «49+ Languages», llms.txt says 50+ for transcription and 40+ for translation, and the pricing page 54+ for transcription and 55+ for translation
the list by name is taken from the languages page and is the primary one; the discrepancy between the round numbers on the home page, in llms.txt and on the pricing page is recorded as it stands. There is no separate list of TRANSLATION languages
source, checked 2026-08-07
Platforms
export formats from the pricing page: DOCX, PDF, TXT, SRT, VTT (the service's llms.txt claims «30+ formats»); subtitles are delivered as SRT/VTT and burnt into the video. Integrations: Zoom, Microsoft Teams, Google Meet, Webex, Zapier, Adobe Premiere; there are a REST API and an MCP server. There are no links to direct publishing to social networks on the pages read
export to social networks is not claimed — the service delivers files, and delivers them into editing software
source, checked 2026-08-07
Pricing
there is no permanent free plan; there is a trial: a «30 min free trial» marked «No credit card required» and «Cancel anytime» (verbatim)
the trial 30 minutes are a one-off sample, not a monthly quota
source, checked 2026-08-07
Pay As You Go — 10 USD an hour of audio without a subscription; Core — 25 USD a month or 275 USD a year (one month free); Advanced — 50 USD a month or 550 USD a year; Pro — 80 USD a month or 880 USD a year; Enterprise — price on request. Extra hours on any subscription cost 10 USD an hour. An extra seat on a subscription costs 25 USD a month
source, checked 2026-08-07
among subscriptions 25 USD a month (the Core plan); the absolute minimum payment is 10 USD for one hour of transcription on Pay As You Go without a subscription
source, checked 2026-08-07
Limits and restrictions
Pay As You Go — hours are bought individually, 5 GB of storage, one user; Core — 5 hours a month, 25 GB; Advanced — 20 hours a month, 50 GB; Pro — 40 hours a month, 100 GB. Hours beyond the package cost 10 USD an hour. Translation is billed separately: verbatim «automated transcription is an extra charge. It is charged at the same rate as your transcription rate», that is, a translated hour is deducted as another hour
no limits on the length of a single file are named on the pricing page or in the terms of use
source, checked 2026-08-07
Restrictions
it is forbidden to upload material that «infringes any intellectual property or other proprietary rights» and that «you do not have a right to upload under any law or under contractual or fiduciary relationships» (verbatim). There is no separate clause in the terms read about the need to obtain consent from the people recorded — the prohibition is framed through rights and lawfulness
SILENCE OF THE SOURCE about the consent of those recorded: there is a rule about third-party rights but no direct requirement to notify the people being recorded
source, checked 2026-08-07
Legal
NO, expressly denied in the privacy policy. Verbatim: «Sonix does not use your content or Customer Data to train our machine learning or generative AI models». It is separately stipulated that when a third-party AI application is connected through OAuth the service passes on «only the data necessary to fulfill the requests you initiate through that application», and that party's policy applies thereafter
the terms of use do not touch on training at all; the denial was found specifically in the privacy policy
source, checked 2026-08-07
no separate opt-out is required, since training on customer data is denied by default. A withdrawal mechanism is described only for third-party connections: verbatim you may «revoke a connector's access at any time through your Sonix account settings or by contacting us»
formally this is not an opt-out from training but a withdrawal of an integration's access; there is no separate opt-out button for training, and none is needed given the stated denial
source, checked 2026-08-07
Catalogue section: all similar See also: catalogue index · scheduler comparison · find by situation · platform restrictions