**Audio Transcription in 99 Languages — Speech to Text | SozAI** Source: https://sozai.app/transcription/ # Audio Transcription in 99 Languages Upload a recording, record a meeting or paste a link. You get text with speaker labels and a timestamp on every word — in the language it was spoken. [Get the App — Free](https://sozai.app/download/) iOS and Android. 30 minutes free, no card. [Try it now — paste a YouTube link](https://sozai.app/youtube-transcript/) What is SozAI audio transcription? SozAI converts audio and video into text in 99 languages, with automatic speaker labels and a timestamp on every word. It runs on iOS, Android and macOS and in the browser: record straight into the app, upload a file up to 500 MB, or paste a YouTube link. Transcripts can be summarised, searched and asked questions, and exported as TXT, PDF, SRT, VTT or JSON. Recordings are stored on our servers in Germany and travel over encrypted connections. Every account includes 30 minutes of transcription, once, at no cost and without a card; YouTube transcripts on the website need no account at all. - 99 languages - Speaker labels - Word-level timestamps - TXT · PDF · SRT · VTT · JSON - EU hosting Specification ## What you can send, and what comes back SozAI transcription — capability sheet v2026.07 | Languages | 99, with automatic detection | | --- | --- | | Audio formats | MP3, M4A, WAV, AAC, FLAC, OGG | | Video formats | MP4, MOV — the spoken track is extracted | | Other sources | YouTube link, in-app recording, screen and meeting audio | | Maximum file size | 500 MB | | Speaker labels | Yes — automatic diarization | | Timestamps | Word level | | Summaries | Yes, in the language of the recording | | Ask the transcript | Yes — question and answer over the recording | | Translation | Yes, 50+ languages including live two-way voice | | Export formats | TXT, PDF, SRT, VTT, JSON | | Platforms | iOS, Android, macOS, web | | Free tier | 30 minutes once, no card | | Data location | Stored on our servers in Germany, encrypted in transit | Coverage ## Languages with a page of their own Each one covers the dialects handled, how the script is written out, and what tends to go wrong in that language specifically. Languages outside this table are still supported. Dedicated language pages 12 | Language | Writing system | Direction | Varieties covered | | --- | --- | --- | --- | | [العربية Arabic](https://sozai.app/transcription/arabic/) | Arabic (abjad) | Right to left | Modern Standard Arabic, Egyptian, Gulf, Levantine, Maghrebi | | [Español Spanish](https://sozai.app/transcription/spanish/) | Latin | Left to right | Castilian, Mexican, Rioplatense, Caribbean, Andean | | [Português Portuguese](https://sozai.app/transcription/portuguese/) | Latin | Left to right | European, Brazilian, African | | [Deutsch German](https://sozai.app/transcription/german/) | Latin | Left to right | Standard German, Austrian, Swiss High German | | [Français French](https://sozai.app/transcription/french/) | Latin | Left to right | Metropolitan, Québécois, Belgian, West African | | [Русский Russian](https://sozai.app/transcription/russian/) | Cyrillic | Left to right | Standard Russian | | [Türkçe Turkish](https://sozai.app/transcription/turkish/) | Latin (Turkish alphabet) | Left to right | Istanbul standard, Anatolian | | [हिन्दी Hindi](https://sozai.app/transcription/hindi/) | Devanagari | Left to right | Standard Hindi, Hinglish code-switching | | [日本語 Japanese](https://sozai.app/transcription/japanese/) | Kanji, Hiragana, Katakana | Left to right | Standard (Tokyo), Kansai | | [한국어 Korean](https://sozai.app/transcription/korean/) | Hangul | Left to right | Seoul standard | | [中文 Chinese](https://sozai.app/transcription/chinese/) | Han (Simplified and Traditional) | Left to right | Mandarin, Cantonese | | [Қазақша Kazakh](https://sozai.app/transcription/kazakh/) | Cyrillic (Latin transition under way) | Left to right | Standard Kazakh, Kazakh-Russian code-switching | The trade ## AI transcription against the alternatives | | SozAI | Typing it yourself | A transcription agency | | --- | --- | --- | --- | | One hour of audio takes | About 6 minutes | 4 to 6 hours | 1 to 3 days | | Speaker labels | Automatic | By hand | Usually included | | Word timestamps | Every word | Rarely | On request | | Languages | 99 | The ones you speak | Depends on the vendor | | Cost per hour | Included in a flat subscription | Your own time | Typically $60–$150 | | Summary and search | Built in | No | No | | Who sees the recording | Nobody — machine only, EU servers | You | A third-party typist | Agency prices are typical published market rates for human transcription, not a quote. Where to start ## Pick the job you have - [Audio to text Upload a file up to 500 MB and get a transcript back.](https://sozai.app/audio-to-text/) - [Video to text Pull the spoken track out of any video file.](https://sozai.app/video-to-text/) - [YouTube transcript Paste a link — no account needed.](https://sozai.app/youtube-transcript/) - [Subtitles Export SRT and VTT with word-level timing.](https://sozai.app/subtitle-generator/) - [Meeting recorder Record the call and get notes with action items.](https://sozai.app/meeting-recorder/) - [Interviews Speaker-separated transcripts you can quote from.](https://sozai.app/interview-transcription/) - [Medical Clinical notes without the evening admin.](https://sozai.app/medical-transcription/) - [Legal Depositions and hearings, on the record.](https://sozai.app/legal-transcription/) - [Translate speech Two-way voice translation in 50+ languages.](https://sozai.app/audio-translator/) Answers ## Audio transcription — your questions The answers people actually ask for before they upload anything. How accurate is AI audio transcription? Accuracy depends on the audio and the language. In our published test, 200 short clips in 10 languages, the word error rate ran from 1.8% for Spanish to 19.6% for Arabic and 50% for Kazakh, on read sentences in quiet rooms. It rises with overlapping speakers, heavy accents, background noise and phone-quality audio. The single change that improves a transcript most is an external microphone — it does more than any setting in the app. How many languages can SozAI transcribe? 99, with automatic language detection from the first seconds of speech. Twelve of them have a dedicated page covering dialect coverage, how the script is written out and what tends to go wrong. Languages without their own page are still supported — set them manually or leave detection on. What audio and video formats are supported? MP3, M4A, WAV, AAC, FLAC and OGG for audio, and MP4 and MOV for video — the spoken track is extracted automatically. Files can be up to 500 MB. You can also paste a YouTube link instead of uploading anything. Does it separate speakers? Yes. Speaker diarization runs on the audio rather than the language, so it labels each voice and timestamps every turn in any supported language. For interviews, depositions and meetings this is what turns a recording into a document you can quote from rather than one long paragraph. Is transcription free? Every account gets 30 minutes of transcription once, at no cost and without a card. YouTube transcripts on the website need no account at all. Beyond that, Premium is a subscription through the App Store or Google Play — from $3.99 a week or $44.99 a year in the US, with up to 700 hours of transcription — and the price does not depend on which language you transcribe. Where are recordings stored? Files are stored on our servers in Germany and travel over encrypted connections. Speech recognition runs at a processing partner; the full list is on our sub-processors page. You can delete a recording and its transcript from the app or the web app at any time. For teams working under GDPR this is usually the first question asked, ahead of accuracy. Can I get subtitles as well as a transcript? Yes. Every transcript can be exported as SRT or VTT with word-level timing, alongside TXT, PDF and JSON. The subtitle files are UTF-8, so non-Latin scripts and right-to-left languages survive the trip into YouTube, Premiere Pro, Final Cut Pro and DaVinci Resolve. Transcribe your first recording 30 minutes once, no card. Record, upload a file or paste a link. [Get the App — Free](https://sozai.app/download/) Free on iOS and Android. No account required for YouTube transcripts. --- This is the markdown twin of https://sozai.app/transcription/ — the same content, without the markup. Published by SozAI (https://sozai.app). Reuse and quotation are allowed with attribution and a link back. Machine-readable index: https://sozai.app/llms.txt · data API: https://sozai.app/api/