Speech to text · தமிழ்
Upload a voice note, a call recording or a video and get Tamil text back. It is free for files up to 2 minutes without an account (3 files a day); longer files need a free account. Download the result as text, Word or subtitles. Spoken Tamil stays spoken in the transcript most of the time, English words come back in Tamil script, and spoken numbers usually turn into digits.
Upload a voice note, call recording or video in Tamil, or record yourself. Free up to 2 minutes a file, 3 files a day, no account.
Upload MP3, M4A, WAV, AAC, FLAC, AMR, a WhatsApp voice note (.opus or .ogg) or an MP4 or MOV video, or record up to 2 minutes from your microphone. Switch between Tamil script, Roman script and an English translation (made by Vyas, our language model, when you open the tab), then copy the text or download TXT or DOCX, plus SRT subtitles for uploaded files. No account needed for files up to 2 minutes, 3 a day.
Drop in an audio or video file and get Tamil text back in Tamil script. It reads MP3, M4A, WAV, AAC, FLAC and AMR audio, WhatsApp voice notes (.opus and .ogg) and MP4 or MOV video, and the file is decoded on our server, so it works the same on an iPhone. You can also record from your microphone. There is no language menu: Valmiki STT, the speech recognition model behind it, works out that the speech is Tamil, including English words in the middle of a sentence.
Copy the transcript, or download it as a TXT file or a Word document. For an uploaded file you also get an SRT subtitle file to load into YouTube Studio or a video editor. Long recordings after sign-up cost $0.00525 per minute of audio, billed by the second, the same for every language.
Real transcripts
Each clip was rendered by Pāṇini TTS and streamed to Valmiki STT twice: as a clean recording and as an 8 kHz phone-quality copy. The transcripts are shown exactly as they came back, errors included. Tested 24 October 2026. These are synthetic clips, not recorded calls.
Same transcript in both phone-quality runs.
One of two clean runs, which wrote the formal காலையில். The other kept the spoken காலையில.
No public benchmark includes Valmiki STT, so we do not publish a table comparing it with other services. Judge it on audio like yours instead. Play each clip above and compare what was said with what came back, word by word. Then upload a recording of your own that is close to what you will really transcribe: the same kind of phone line or microphone, the same speakers, the same mix of Tamil and English. Count the words that changed meaning, not the spelling variants, and try the 8 kHz case if your audio comes from calls.
About Tamil
Spoken Tamil is not the Tamil of newspapers. People say காலையில rather than காலையில், and use English words like appointment and doctor. Valmiki STT kept the spoken form காலையில in three of four runs of our test clip, and wrote the English words in Tamil script (டாக்டர், அப்பாயின்ட்மென்ட்).
Spoken numbers came back as digits: பத்து மணிக்கு (at ten o'clock) was written 10 மணிக்கு in every run. It also made the same small slip every time: டாக்டர பார்க்க (to see the doctor) came back as டாக்டர் பார்க். Check your own audio before you rely on exact wording.
Dialects and mixing
The gap between written and spoken Tamil is wide, and a transcript has to pick one. Valmiki STT mostly writes what was said (காலையில, not காலையில்), which is what you want for a voice note or a call, but it sometimes switches to the written form within the same recording. If you need consistent formal Tamil, edit the text afterwards.
Tanglish, Tamil mixed with English words or whole English phrases, is handled without a setting: the English words land in Tamil script inside the sentence. We tested Chennai-style speech only. Madurai, Kongu and Tirunelveli speech, and Sri Lankan Tamil from Jaffna or the diaspora in Canada and the UK, have their own words and sounds; try a recording of your own before you rely on it.
Use cases
Voice notes from family, from a tuition teacher or from a shop confirming an order. Upload the .opus file and read the Tamil, or open the English tab for someone outside Tamil Nadu.
Tamil cooking channels, tech reviews, sermons and talk shows. For your own videos, download the MP4 from YouTube Studio and take the SRT file as a start on Tamil captions.
Interviews for Tamil journalism and research, and recorded meetings in Chennai offices, where Tamil and English trade places every few words. Long recordings go in parts of up to 10 minutes after a free sign-up.
Clinic and service calls where the caller gives a date and a time; Valmiki writes the time as digits, which makes it easy to find in the text. Teams with many calls send them to the API.
Voice typing
Voice typing is a different job from transcribing a file: you speak and the words appear as you go. Sonex Labs has no voice-typing keyboard or app, so for that, use the tools already on your phone or computer.
1. Android phone (Gboard)
Install Gboard, add Tamil (தமிழ்) under Languages in its settings, then tap the microphone on the keyboard. The words come out in Tamil script.
2. iPhone
Add the Tamil keyboard in Settings, then General, Keyboard, Keyboards. If your iPhone offers dictation for Tamil, a microphone key appears on that keyboard.
3. WhatsApp
To type a message by voice, tap the microphone on your keyboard, not the voice-note button in WhatsApp; the keyboard writes the words into the chat box. WhatsApp can also show a transcript under a received voice message in some languages (Settings, then Chats, then Voice message transcripts). If Tamil is not offered there, save the voice note and upload it above.
4. Microsoft Word
Word's Dictate lists Tamil (India) as a preview language, which Microsoft says may be less accurate and add less punctuation. Pick it in Dictate's settings and speak.
5. Windows
Windows 11 voice typing lists Tamil (India). Switch the keyboard input to Tamil (Windows key + Space), press Windows key + H in a text box and speak.
For recordings longer than a minute, use the uploader above.
API
Send a recording to POST /v1/transcribe and read the Tamil text from the reply. There is no language parameter. $0.00525 per minute of audio, billed by the second.
1# pip install requests2import requests34with open("tamil-call.wav", "rb") as audio:5 r = requests.post(6 "https://api.sonexlabs.com/v1/transcribe",7 headers={8 "Authorization": "Bearer YOUR_API_KEY",9 "Content-Type": "audio/wav", # audio/mpeg, audio/ogg (WhatsApp notes), audio/mp4 ...10 },11 data=audio,12 )13r.raise_for_status()14result = r.json()15print(result["text"]) # வணக்கம், நாளைக்கு காலையில 10 மணிக்கு ...16print(result["duration_seconds"], "seconds of audio")About ₹0.50 a minute (₹29.93 an hour of audio), at exchange rates from open.er-api.com, 25 October 2026. Live streaming, limits and error codes are on the speech to text API page, and plans are on pricing.
FAQ
Yes, for short files. The uploader on this page turns Tamil audio into text with no account for files up to 2 minutes, 3 files a day. For longer files, sign up free and transcribe files of up to 10 minutes and 10 MB each in the dashboard or through the API; that costs $0.00525 a minute, billed by the second, so a 10-minute recording costs about 5.25 US cents.
There is no app to install. The uploader works in the browser on a phone too: open this page on Android or iPhone, pick a voice note, recording or video from your files, and read the Tamil text. For typing by voice as you speak, use your phone keyboard's microphone, as described in the voice-typing steps above.
Yes. It runs in a current browser: Chrome, Edge, Firefox or Safari on a Windows PC or a Mac, and Safari or Chrome on an iPhone or Android phone. Files are decoded on our server, so an iPhone voice memo, a WhatsApp .opus note or a MOV video works the same as on a PC. There is nothing to install.
Long-press the voice note in WhatsApp and share or export it to your files; on Android, received voice notes are also kept as .opus files in WhatsApp's media folder. Upload the .opus or .ogg file here and the Tamil text appears. Notes up to 2 minutes need no account.
Upload the audio, then open the English tab above the transcript. Vyas, our language model, translates the Tamil transcript when you open the tab, and you can copy it or download it like the transcript. Check names, amounts and dates against the Tamil text, because a word misheard in Tamil carries into the English.
Yes. The uploader's Roman script tab shows the transcript in English letters, the way Tanglish is typed on WhatsApp. The API returns Tamil script, as it did in every run of our sample, so convert the text yourself if your system needs Latin letters.
On our test sentence every run kept the meaning and wrote the time as 10. Every run also made the same slip, டாக்டர பார்க்க heard as டாக்டர் பார்க், and one clean run switched காலையில to the formal காலையில். That is one sentence, not a benchmark; test a few recordings like yours.
Without an account, 2 minutes per file and 3 files a day. With a free account, up to 10 minutes and 10 MB per audio file in the dashboard or the API, billed by the second. For anything longer, split the recording into parts of 10 minutes or less and send each one; the API takes the same limit.
Usually digits. In every run of our test clip, பத்து மணிக்கு (at ten o'clock) came back as 10 மணிக்கு. If you need the spoken form, change the digits back after transcription.
Related
Need the reverse? Turn Tamil text into speech with Pāṇini TTS, free to try without an account.
Answering Tamil phone calls rather than transcribing them? See Tamil voice agents.
Free for files up to 2 minutes without an account. Longer files, with a free account: $0.00525 per minute of audio, billed by the second.