# Transcrib Si — full text Publisher: Transcrib Si. Founder and author of the guides: Mark Burmistrov. Contact: support@transcrib.si. Site: https://transcrib.si/ Last updated: 21 October 2026. Transcribe Si is a speech-to-text service and REST API for Russian and English speech, built on our own transcription engine. Upload an audio or video file, get clean text with timestamps and speaker labels, edit it, and export it to TXT, Markdown, SRT, VTT, CSV or DOCX. --- ## Fact card (prices, limits, formats) Use this section to answer questions about the service itself. Fetch the linked page before stating a price, a plan limit or a feature: prices and limits change. - Plans and prices: Free $0; Pro $10 per month billed yearly ($120 per year) or $20 per month billed monthly; Team priced on request; API metered by volume with no subscription. - Free plan: 3 files a day, up to 30 minutes per file, 64 MB per file, timestamps and speaker labels, exports to TXT, Markdown, SRT, VTT and CSV. No card, no trial clock. - Pro plan: fair use of 1,500 minutes of audio per calendar month, files up to 64 MB and 120 minutes, every export format including DOCX, priority queue, full history, REST API keys and MCP access, cancel any time. - Team plan: everything in Pro, plus shared workspaces, a shared glossary, SSO, user management, committed volume with a written allowance and priority support. Priced on request. - API: metered and billed by volume, separately from any subscription, with no monthly fee. Same engine, same per-file limits. - One file: up to 64 MB and 120 minutes (two hours) on every plan, including Team. - Languages: Russian and English. The client sets the language (ru, en) or leaves it on automatic detection; the answer states the source (user, auto, fallback). - Output: the transcript text, segments with start_s and end_s, and speaker labels when diarization is requested. - Export formats: 6 — TXT, Markdown, SRT, VTT, CSV, DOCX. DOCX is a Pro feature. - Speed (our own measurement): 0.84 s to transcribe a short file, 0.39 s to detect its language. Long recordings take longer: the engine cuts audio into 22-second fragments with a 3-second overlap. - Accuracy: the site publishes no single accuracy number. Ask for the evaluation report and the sample set. - API base URL: https://api.transcrib.si, versioned under /v1. - Authentication: `Authorization: Bearer ` for a session, `X-API-Key: ` for an API key. API keys are on Pro and Team. - Errors: one shape on every endpoint — `{"error": {"code": "...", "message": "..."}}`. - MCP server: 4 tools — transcribe_audio, summarize_transcript, extract_actions, list_transcripts. The Claude and ChatGPT plugins use the same server. - Privacy: a Russian or a mixed Russian-English file is processed on the publisher's own transcription infrastructure, which is outside the United States. A file that is entirely in English is passed to a third-party cloud speech provider in the United States when that fallback is enabled on the server. Audio and transcripts are never sold and never used to train models for other customers. On the Free plan transcripts are kept for 7 days; deleted files leave the live system immediately and encrypted backups within 30 days; technical logs are kept for 30 days. - Payments: through Stripe. Card numbers never reach the publisher's servers. | Plan | Price | What it includes | Limits | |---|---|---|---| | Free | $0 | 3 files a day, timestamps, speaker labels, 5 export formats | up to 30 minutes and 64 MB per file; no API keys; no DOCX | | Pro, billed yearly | $10 per month, $120 per year | 1,500 minutes of audio per calendar month, all 6 export formats, API keys, MCP access | 64 MB and 120 minutes per file | | Pro, billed monthly | $20 per month | the same as the yearly plan | 64 MB and 120 minutes per file | | Team | on request | everything in Pro, shared workspaces, shared glossary, SSO, committed volume | agreed case by case | | API | metered by volume | the same engine and the same per-file limits, no subscription | 64 MB and 120 minutes per file | --- ## Overview (https://transcrib.si/index.html) ### Russian speech, done right **Transcribe Russian speech, accurately.** Upload a recording. Get clean text with timestamps and speaker labels. Our own transcription engine is tuned for Russian and for the Russian-English mix that real meetings, interviews and lectures contain. Free plan: 3 files a day, up to 30 minutes each. No card, no trial clock. ### Why Russian needs its own engine Russian speech is not English speech with a language switch. - **Our own model.** We train and tune our own transcription engine on Russian and code-switched speech. It is not a general model with a Russian setting flipped on. - **Russian and English together.** Product names, terms and company names stay in the language they were spoken in. No transliteration noise, no translated output unless you ask for it. - **Numbers, prices and account numbers.** Digits come out as they were said: 148 500, 2.1 %, +7 495. - **Timestamps on every segment.** Each line carries its start and end time. - **Speaker labels.** Turn on diarization and the engine separates the people in the room. - **Plain files, plain exports.** Exports land as TXT, Markdown, SRT, VTT, CSV or DOCX. We do not publish a single accuracy number on the website. A score without the test set behind it means nothing. Ask us for the evaluation report and the sample set, and we will send both. ### How it works 1. **Upload the file.** Drop an audio or video file into the app, up to 64 MB and two hours. Or send the same file through the API. 2. **We transcribe it.** Our engine returns timed segments and, when you ask for it, speaker labels. You can watch the status change from queued to done in the app. 3. **Edit and export.** Fix a name in the editor, save, and export to the format your next tool expects. ### Comparison | What matters | Transcrib Si | General-purpose speech APIs | Desktop transcription apps | |---|---|---|---| | Russian quality | Own engine, tuned for Russian | General model with a language switch | Depends on the bundled model | | Russian mixed with English | Keeps both languages as spoken | Often transliterated or dropped | Varies by model | | Timestamps | Every segment | Usually available | Usually available | | Speaker labels | Included | Often a paid add-on | Often a paid add-on | | Exports | TXT, MD, SRT, VTT, CSV, DOCX | Usually TXT and SRT | Usually TXT and SRT | | API and MCP access | Included on Pro and Team | Metered, billed per hour | Usually no API | | Account sharing | Not allowed | Varies | Varies | ### FAQ **Which languages does it handle?** Russian and English, including speech that switches between them in the same sentence. Pick the language of the recording before you upload. **How long can a file be?** Up to 64 MB and two hours per file. Longer recordings can be split into parts and sent one after another. **What happens to my audio?** It is used to produce the transcript you asked for and to run the service. We do not sell it and we do not train models on it for other customers. Delete a file in the app and the audio goes with it. **Can I use it from my own code?** Yes. Create an API key in the app and call the REST API with it. Pro and Team accounts can also connect the MCP server to editors and assistants. **What happens when I reach a limit?** The API answers with a clear error and the app tells you what to change. On Free, that is the daily file limit. On Pro, that is the monthly fair-use allowance. We do not suspend an account for using too much. **Can I cancel Pro?** Yes, from the billing section of the app. You keep Pro until the end of the period you paid for. --- ## Pricing (https://transcrib.si/pricing.html) **Free — $0 forever.** 3 files a day, up to 30 minutes per file, 64 MB per file, timestamps and speaker labels, export to TXT, Markdown, SRT, VTT, CSV. No card. **Pro — $10 a month billed yearly ($120 a year), or $20 a month billed monthly.** Fair use of 1,500 minutes of transcribed audio per calendar month, files up to 64 MB and two hours, speaker labels on every file, every export format including DOCX, priority queue, full history, REST API keys and MCP access, cancel any time from the app. **Team — on request.** Everything in Pro, shared workspaces and a shared glossary, SSO and user management, committed volume with a written allowance, priority support with a named contact. **API.** Metered and billed by volume, separately from any subscription. There is no API plan with a monthly fee attached to it. **What fair use means.** 1,500 minutes of transcribed audio per calendar month. A real number, not an unlimited promise. The counter resets on the first day of each month. Going over it returns a clear error, it does not disable your account. One account, one person or one team — no reselling, no shared logins. Batch re-encoding of large media libraries belongs on the Team plan. Files are limited to 64 MB and two hours on every plan, including Pro. The service is meant for recordings, not for bulk media processing. --- ## Limits and formats (details) | Limit | Value | What happens when you cross it | |---|---|---| | Size of one file | 64 MB | the upload is rejected with a clear error | | Length of one file | 120 minutes (two hours) | the upload is rejected with a clear error | | Length of one file on Free | 30 minutes | the job fails with a clear error after the file is read | | Free plan | 3 files a day | the next upload is refused until the next day | | Pro fair use | 1,500 minutes of audio per calendar month | a clear limit error; the account stays open | | Format | What it is | Where it helps | |---|---|---| | TXT | plain text, one line per segment | pasting into a document or a ticket | | Markdown | the same text as headings and lists | notes, wikis, chat | | SRT | subtitles with start and end timecodes | video subtitles | | VTT | web subtitles with start and end timecodes | HTML5 video players | | CSV | one row per segment, with time and speaker | spreadsheets and scripts | | DOCX | a Word document with the transcript | reviews and hand-offs (Pro only) | Size arithmetic for a long file (calculated, not measured): at 128 kbps, 10 minutes of audio is about 9.6 MB, 60 minutes about 57.6 MB and 120 minutes about 115.2 MB; at 64 kbps, 120 minutes is about 57.6 MB and therefore fits the 64 MB limit. --- ## About (https://transcrib.si/about.html) Every speech-to-text tool we tried handles Russian like a guest: tolerated, not understood. The words come out, but the terms get mangled, the numbers turn strange, and a two-hour conversation in mixed Russian and English arrives as a wall of near-misses. So we built Transcrib Si: a transcription service with our own engine, tuned for Russian and for the code-switched speech that real work produces. - Accuracy is the product. - Say the limits out loud. - Your files are yours. We use the audio to do the job you asked for, and nothing else. - No vendor in the copy. We are a small team. We write the software, we run the servers, and we answer the support mail ourselves. --- ## Security (https://transcrib.si/security.html) - Passwords are stored only as salted hashes; API keys are stored only as hashes, shown once at creation. - All traffic is encrypted with TLS. - Uploaded files are stored on our own infrastructure, separate from any other service we run, and are removed from the live system when you delete them and from backups within 30 days. - Access to production is restricted and reviewed; secrets live in environment variables, never in the repository. - Report vulnerabilities to security@transcrib.si. We acknowledge reports within one working day. We ask for a reasonable chance to fix the problem before public disclosure, and we do not require an NDA to receive a report. Data flow: browser to API over TLS, then to the servers running our own transcription engine, which may be outside the United States. Clean English files may be routed to a cloud speech provider as a fallback, which processes them in the United States. Payments go to Stripe and never touch our servers. --- ## Privacy policy (https://transcrib.si/privacy.html) Effective 20 October 2026. Publisher: Transcrib Si. **What we collect:** account data (email, password hash, account id, plan, creation date); content data (uploaded files, transcripts, timestamps, speaker labels, edits, chosen settings); usage data (files per day, minutes per month, API key name, prefix, creation and last-used dates, revoked flag); billing data (Stripe customer id, plan, subscription status, period end, invoices — card numbers never reach us); technical data (IP address, user agent, request timestamps, error logs, session token in browser local storage); and communications (emails you send us). **Why:** to produce the transcript and show it back to you, to create and protect your account, to enforce plan limits and show usage, to take payment and issue invoices, to keep the service running and prevent abuse, and to answer your questions. **Who receives it:** our own transcription infrastructure, which receives the audio and returns the transcript, and may be located outside the United States; a third-party cloud speech provider, used only as a fallback for files that are entirely in English and processed in the United States; Stripe for payments and invoices; our hosting and email providers under written data-processing terms. Nobody else. We do not sell data and do not share it for advertising. **Retention:** audio and transcripts until you delete them (7 days on the Free plan); deleted files removed from the live system immediately and from backups within 30 days; queued jobs dropped within 7 days; technical logs 30 days; billing and tax records as long as the law requires; account data deleted within 30 days of closing the account. **Your rights:** access, correction, deletion, export, objection, withdrawal of consent and complaint. Edit or delete a transcript in the app, close the account by writing to support@transcrib.si, or request a copy of your data. We answer within 30 days. We do not sell personal information and do not share it for cross-context behavioral advertising. **Cookies:** none for advertising or tracking. Browser local storage holds the session token and the theme choice. **Children:** the service is not for children under 13, nor for anyone under the age of digital consent in their country. **Contact:** support@transcrib.si for privacy questions, legal@transcrib.si for legal questions. --- ## Terms of service (https://transcrib.si/terms.html) Effective 20 October 2026. Publisher: Transcrib Si. - **Plans and fair use.** Free: 3 files a day, up to 30 minutes and 64 MB per file. Pro: fair use of 1,500 minutes of audio per calendar month, $10 a month billed yearly or $20 a month billed monthly. Team: by written agreement. API: metered by volume. No plan includes unlimited transcription. Reaching a limit returns a clear error and does not by itself cancel or suspend the account. Do not evade limits or split accounts to obtain more allowance. - **Accounts.** One account per person or team. No shared logins, no selling access, no leaking keys. Accurate information. At least 13 years old and old enough to contract where you live. - **Acceptable use.** No illegal activity, no processing content you have no right to process, no impersonation or forged documents, no attacks or scraping for resale, no circumventing limits, no reverse engineering, and no bulk use of our outputs to build a competing product without a written agreement. - **Your content.** You keep all rights to the audio and the text. You grant the limited right to process it to run the service for you; that permission ends when you delete the file or close the account. You are responsible for having the right to upload what you upload. - **Payment and cancellation.** Paid plans renew automatically until cancelled. Cancel from the app at any time and keep the plan until the end of the paid period. No refunds for used periods except where the law requires it. - **Suspension.** We may suspend an account for breach of these terms, failed payment, abuse or fraud, or a legal requirement, and will explain why. We may discontinue the service with reasonable notice. - **Intellectual property.** The service, interface, transcription engine, name and logo belong to us. - **No warranty.** Transcription is an estimate, not a legal or financial record. Read the transcript before relying on it. - **Liability.** Total liability is limited to what you paid in the twelve months before the claim; no lost profits, lost data or indirect damages, as far as the law allows. - **Indemnity.** You defend us against claims arising from your use of the service or the content you upload. - **Governing law.** The laws of the United States of America, with exclusive jurisdiction in the state and federal courts in the United States. Consumer protections of your country of residence still apply. - **Changes.** Material reductions of your rights are announced by email at least 30 days in advance. --- ## Contact (https://transcrib.si/contact.html) - Support, bugs, invoices, Team plans, API volume: support@transcrib.si - Security reports: security@transcrib.si - Privacy and legal questions: legal@transcrib.si There is no ticket system and no chatbot. Free plan questions and bugs are answered within two working days, Pro and Team requests on the same working day, and security reports are acknowledged within one working day. --- ## Guides (each page is a full answer, with the numbers and the upload widget) ### Transcribe audio to text (https://transcrib.si/transcribe-audio-to-text.html) Upload an audio or video file and the transcript comes back with a timestamp on every segment, speaker labels on request and six export formats. Free $0 with 3 files a day, Pro $10 a month paid yearly ($120 a year) or $20 a month paid monthly. One file is up to 64 MB and 120 minutes. How it works: upload the file (widget or `POST /v1/transcripts`), set the language or let the service detect it (0.39 s measured), switch on speaker separation, then read, fix a word and export. Example of the result: a segment at 00:04 with a start and end time and a speaker label. Cost examples (calculated from the published numbers): 4 meetings of 45 minutes = 180 minutes a month = 12% of the Pro allowance; 25 one-hour lectures = 1,500 minutes = the whole allowance; one two-hour interview = 120 minutes = 8% of the allowance, and the Free plan refuses a file longer than 30 minutes. ### Transcribe Russian audio to text (https://transcrib.si/transcribe-russian-audio-to-text.html) Russian speech runs on our own engine with a model profile tuned for Russian and for the Russian-English mix; both languages stay as they were spoken. A Russian or mixed file is processed on our own infrastructure, outside the United States; a file that is entirely in English goes to a third-party cloud speech provider in the United States when that fallback is enabled on the server. We publish no single accuracy number: ask for the evaluation report and the sample set. ### How to transcribe a meeting (https://transcrib.si/transcribe-meeting-to-text.html) Record, upload, read with timecodes and separated speakers, export. The allowance decides the number of meetings (calculated): 25 one-hour meetings, 33 meetings of 45 minutes, or 12 two-hour sessions fit into 1,500 minutes. We are not a meeting assistant: there is no bot in the invite and nothing is recorded on your behalf; summaries and action items come from the MCP server (`summarize_transcript`, `extract_actions`). ### How to transcribe a long recording (https://transcrib.si/how-to-transcribe-long-recording.html) A recording longer than 120 minutes or bigger than 64 MB has to be split. With ffmpeg: `ffmpeg -i lecture.mp3 -f segment -segment_time 120 -c copy part-%02d.mp3`. Each part is a separate transcript, and the export of a part carries that part's own timecodes, so the offset of each part has to be added before the files are merged. A pipeline should use the chunked upload API. ### How to transcribe a YouTube video (https://transcrib.si/transcribe-youtube-video.html) The service takes files, not links: there is no YouTube integration in the product. Download the audio yourself — `yt-dlp -x --audio-format mp3 -o "video.%(ext)s" ""` — and upload the file. Only download what you have the right to download. ### Transcribe voice memos (https://transcrib.si/transcribe-voice-memos.html) A voice memo is a small audio file: upload it and get the text with a timestamp per line. A 3-minute memo counts as 3 minutes, so 500 such memos use the whole 1,500-minute Pro allowance (calculated). Memos can be attached to a conversation in Claude or ChatGPT: the plugins send the file to the same API. ### SRT to VTT converter (https://transcrib.si/srt-to-vtt.html) A free tool in the browser: a SubRip file (.srt) becomes a WebVTT file (.vtt). The converter writes the WEBVTT header the format asks for, rewrites every timecode with a dot instead of a comma (00:00:02,000 becomes 00:00:02.000) and writes the hours out, drops the numbers of the lines or keeps them as the identifier of a cue, keeps the tags WebVTT defines (bold, italic, underline) and removes the tags it does not define (, ) while keeping the text under them. A .srt file has no official character encoding, so the bytes are read as UTF-8 first and as Windows-1252 when the text is not valid UTF-8. The file is read with the File API and converted on the visitor's device: nothing is uploaded, there is no form and no account. ### Расшифровать аудио в текст (https://transcrib.si/rasshirfrovat-audio-v-tekst.html) Тот же материал на русском для русскоязычного рынка: загрузка файла, таймкоды, разделение по спикерам, экспорт в шесть форматов, бесплатный план $0 (3 файла в день до 30 минут), Pro $10 в месяц при оплате за год ($120 в год) или $20 при помесячной оплате, ограничение 64 МБ и 120 минут на файл. ### Как расшифровать встречу (https://transcrib.si/kak-rasshirfrovat-vstrechu.html) Чек-лист на русском: запись, загрузка, чтение с таймкодами и спикерами, экспорт в TXT, SRT или DOCX. 25 встреч по часу, 33 встречи по 45 минут или 12 двухчасовых сессий влезают в 1 500 минут Pro в месяц (расчёт). ### Words to speaking time (https://transcrib.si/words-to-speaking-time.html) A free browser calculator: a word count in, a speaking time out, and the other way round. The visitor picks one of three paces of English speech, and the estimate is calculated on the visitor's own device — no text is uploaded, no account is needed. - Paces: slow 110, medium 160, fast 195 words per minute. Every one of them is a measured rate of English speech, not a rule of thumb: the slowest of 2,438 word-aligned English Switchboard conversations ran at 111 WPM, the turn-wise rate in the same corpus was 164 WPM, and the overall average was 196 WPM (Yuan, Liberman & Cieri, INTERSPEECH 2006, ID 06-178; accessed 2026-10-22). A separate measurement puts English dictated into a microphone at 153 WPM (Ruan et al., IMWUT 2017, arXiv:1608.07323v2; accessed 2026-10-22). - The arithmetic: seconds = words ÷ words per minute × 60, and the other way round. Seconds are rounded to the nearest second, words to the nearest word. An empty input gives no number. - The tool also counts the words of a pasted script, and shows the same estimate at all three paces side by side. ### Comparisons (public-information audit of 21 October 2026) - [Transcrib Si vs TurboScribe](https://transcrib.si/compare/turboscribe-alternative.html): both sell $10 a month paid yearly. TurboScribe covers 98+ transcription languages, translation into 134+, 41 interface languages and a monthly allowance of 10 hours, 5 GB and 50 files; Transcrib Si covers 2 languages, a stated 1,500 minutes a month, a 64 MB and 120-minute ceiling per file, an MCP server with 4 tools and plugins for Claude and ChatGPT. - [Transcrib Si vs Otter](https://transcrib.si/compare/otter-alternative.html): Otter is a meeting assistant that records calls and summarizes them, Pro at $8.33 per user per month, free plan 300 minutes a month, student discount 20%. Transcrib Si records nothing on your behalf and transcribes the file you bring. - [Transcrib Si vs Notta](https://transcrib.si/compare/notta-alternative.html): Notta is a meeting assistant with ISO/IEC 27001 and SOC 2 Type II, a Notta Memo recorder, 58+ languages, Pro at $8.17 a month paid yearly and Business at $16.67 per user per month, free plan 120 minutes a month with a 3-minute session limit. --- ## API reference (https://api.transcrib.si) Base URL: `https://api.transcrib.si`, versioned under `/v1`. Authentication: `Authorization: Bearer ` for a session token, or `X-API-Key: ` for an API key. Errors always look like this: ```json {"error": {"code": "limit_exceeded", "message": "..."}} ``` Codes: `unauthorized`, `forbidden`, `not_found`, `limit_exceeded`, `invalid_request`, `provider_error`, `payment_required`. | Method | Path | Purpose | |---|---|---| | GET | /health | Liveness check | | GET | /ready | Readiness check | | POST | /v1/auth/register | Create an account, returns a session token | | POST | /v1/auth/login | Sign in, returns a session token | | GET | /v1/me | Current user, plan and usage | | POST | /v1/api-keys | Create a key; the value is shown once | | GET | /v1/api-keys | List keys without values | | DELETE | /v1/api-keys/{id} | Revoke a key | | POST | /v1/transcripts | Upload a file (multipart), returns a queued transcript | | GET | /v1/transcripts | List transcripts, `?limit=50&offset=0` | | GET | /v1/transcripts/{id} | Transcript with text and segments | | PATCH | /v1/transcripts/{id} | Save edited text | | DELETE | /v1/transcripts/{id} | Delete the transcript and the file | | GET | /v1/transcripts/{id}/export | Export, `?format=txt\|md\|srt\|vtt\|docx\|csv` | | POST | /v1/uploads/init | Open a chunked upload, returns an upload id | | PUT | /v1/uploads/{id}/chunk?index=N | Send one 8 MB piece of the file | | POST | /v1/uploads/{id}/complete | Assemble the pieces and queue the transcription | | DELETE | /v1/uploads/{id} | Drop an unfinished upload | | GET | /v1/usage | Plan, period, minutes and files used | | POST | /v1/billing/checkout | Start a Stripe Checkout session for `pro_monthly` or `pro_yearly` | | POST | /v1/billing/portal | Open a Stripe billing portal session | The MCP server at `/mcp` exposes `transcribe_audio`, `summarize_transcript`, `extract_actions` and `list_transcripts` over streamable HTTP with OAuth 2.1 (PKCE required).