Frequently asked questions

One place between the recording and the finished work: a transcript with word-level timing and speakers told apart, AI minutes and action items that link back to the moment they were said, subtitles with translation, and a scored review of sales and support calls.

One hundred, detected automatically — every language the speech engine knows. Thirty-two of them, including English, Polish, German, French, Spanish, Chinese, Japanese, Korean and Arabic, also get word-level timestamps, which makes subtitle splitting land exactly on the word. In the rest the timings come from the speech engine alone and can be off by a fraction of a second; everything else works the same. AI documents can be written in any of the hundred, whatever the recording language.

Speakers are separated automatically. You can move a single line to another speaker, merge two speakers the engine split by mistake, and rename anyone once so the name appears everywhere. Very short interjections (“yes”, “ok”) are the most common thing to correct.

Each decision, task and objection links to the exact sentence, so you can check it with one click. Items that point to no moment in the recording, names nobody said and deadlines nobody mentioned are removed automatically, and the document tells you how many.

Input: MP3, WAV, M4A, MP4, MKV, OGG, FLAC, WMA, OPUS, OGA, AMR, 3GP, WEBM, MOV, AAC, M4V and more. Output: DOCX, PDF, Markdown, TXT, CSV, SRT, WebVTT, files for NVivo, MAXQDA and ATLAS.ti, or copy to clipboard. Large files can be made smaller in your browser before sending.

SRT and WebVTT with readability rules — up to 42 characters per line and two lines per subtitle (less for Chinese and Japanese), with timing aimed at 17 characters per second. Subtitles that are still too fast to read are counted and listed before you download. Up to four translated languages per recording; sentences the model was unsure of, or that changed in the transcript, are marked.

Recordings and transcripts are encrypted in storage and processed only in the European Union. AI documents and translations are made from the text, so the audio never leaves storage for them.

Yes. Start from the built-in sales or support scorecard and edit the criteria, or write your own set. Each criterion comes back with a verdict, a justification and the moment in the call it was based on.

Yes, though speaker separation gets harder on noisy or phone recordings. When the engine splits one person in two, you can merge them in one click.

Yes, on every plan. The export window has files for MAXQDA and ATLAS.ti (Word or WebVTT), for NVivo on Windows and for NVivo on Mac, with speakers and timestamps laid out the way each program's documentation describes. Our help page “Import into NVivo, MAXQDA and ATLAS.ti” walks through each import step by step.

Yes. Clean verbatim leaves out filler sounds such as “um” and “uh”, and stuttered repeats such as “the the” — in a view in the editor and in the files you export. Filler sounds are recognised in 28 recording languages; in the others only the repeats are removed. The transcript itself stays word for word: the clean view is read-only, and you can switch back at any time.

Yes. On the home page and on the pages for file formats and recording languages you can upload one recording of up to 5 minutes without signing up and read its transcript with speakers and timestamps. There is one free try a day per internet connection. Without an account we delete the file and the transcript after 24 hours; if you create a free account in the same browser within that time, they move to your account.

Yes, on every plan. In Settings you choose when to delete just the audio, keeping the transcript, and when to delete everything — after 1, 7, 30, 90 or 365 days from the upload. Each recording can have its own setting. We email you 3 days before anything is deleted, so you can change your mind.

Yes, on every plan. “Search all recordings” finds a word or phrase in every transcript in your account — with your corrections and speaker names — ignoring capitals and accents unless you tick Match case, and opens the recording at that moment. “Ask your recordings” answers a question about one recording, the ones you choose or all of them, using only your transcripts and writing in the language of your question. Every statement links to the moment it came from, statements without a link are removed, and if the answer is not in your recordings, you are told so. Search uses no AI jobs; each question uses 1. When the AI server has been idle, the first answer can take a minute or two.

Yes, on every plan. In the transcript, choose Share and create a link. You decide whether the person can listen to the audio (on by default), which AI documents they see, whether they can comment and whether they can download the transcript as DOCX or PDF — the audio file itself is never offered for download. The link works for 7 days, 30 days or with no end date, and you can turn it off at any time; it stops working at once. Reading needs no account, and a comment needs a free account. Each comment is tied to a line and a moment of the recording, signed with the name its author chose, never with an email address.

Yes — up to 20 files at once, on every plan. Each file is checked against your plan before anything is sent: a file longer than your plan allows is not uploaded, and you see why. Language and number of speakers apply to the whole batch, and you can change them for a single file. Files are uploaded and transcribed in the order you chose them. On Free, files beyond the 3 transcripts a day are uploaded but wait as “Uploaded, not started” until the counter resets at midnight UTC. If the connection drops while a large file is uploading, you can resume it, and only the missing parts are sent.

Yes. In your recordings list, press Record, allow the microphone and talk — you see the time and the sound level, and you can pause and stop. The sound is sent to us in parts while you record, so if the page closes or the connection drops, what was recorded until then is kept. After Stop, you name the recording and transcription starts. It records what your microphone hears — not the sound of other tabs or apps such as Zoom, Teams or Meet — and no bot joins your meetings. On a phone, keep the screen on while recording. A recording can be as long as the longest file your plan takes.

Yes, from the website — there is nothing to download from an app store. On Android, open sonafact.com in Chrome and tap “Install app” in the menu. SonaFact then also appears in the Share menu, so a voice message from WhatsApp or Telegram uploads to SonaFact without saving it first. On iPhone, open sonafact.com in Safari, tap “Share” and then “Add to Home Screen”. iOS does not let websites appear in the Share menu, so on iPhone you choose the file inside SonaFact.

Yes. Open “Check a quote” in a recording or on your list of recordings, paste the quote and choose one recording or all of them. SonaFact finds where it was said and shows which words were changed, added or left out, and which words the engine was unsure about there — and you can play that moment. If nothing is close enough, you see “Not found” and up to three closest passages. It is text matching, not AI: it is on every plan, uses no AI jobs, and your quote is not saved.

Yes. Turn on Highlight above the transcript, select text and pick one of four colours; you can add a note. The Highlights tab lists them, and you can download just the highlights as DOCX or TXT. A clip is a stretch of the recording of up to 10 minutes: share it by link, and the person sees and hears only that part. Highlights and links to clips are on every plan; downloading the audio of a clip starts with the Starter plan.

The transcription simply does not start, and the message tells you how much you have used and the date the counter resets. The same goes for the daily AI allowance. Nothing is ever charged on top of the plan you chose.

Yearly is one charge instead of twelve, which costs us less to process. Pro is $10 per month billed yearly, $120 in total, against $15 month to month.

Volume and tools. Starter covers 20 hours a month with files up to 3 hours and adds AI minutes and action items to the summaries. Pro covers 60 hours with files up to 6 hours, adds chapters and the sales and support call reviews with their scorecards, translates subtitles into four languages instead of one, and lets you write your own scorecard criteria. Studio covers 300 hours with files up to 10 hours and raises every daily allowance. What none of them change is the transcript itself — the same engine, the same speaker separation and the same exports in all four.

Yes, on the Plan page in your account. Moving up takes effect at once and you pay only the difference for the rest of the period. Moving down or cancelling takes effect at the end of the period you have already paid for, and the account then moves to the free plan. Either way you keep every transcript and recording you already have — only the monthly hours and the maximum file length change.

Yes. Paid plans are sold by Dodo Payments, our reseller and merchant of record: it works out the tax at checkout, takes your VAT ID there and issues the invoice. Your invoices are on the Plan page in your account, ready to download.

Write to support@sonafact.com with a short description of what you are working on and we will look at it case by case.

Card, Apple Pay or Google Pay (Google Pay is not available in India). Depending on your country, checkout also offers local methods that work for a subscription: UPI and RuPay in India, Pix in Brazil, Kakao Pay and Naver Pay in South Korea, SEPA Direct Debit in the eurozone, Satispay in Italy, Cash App Pay in the US, and Klarna for amounts over $50 in the US and parts of Europe. Checkout shows only the methods available for your country and currency. In India, your bank notifies you before each renewal and takes the payment about 48 hours later, under Reserve Bank of India rules.