Speech to Text

Turn Khmer Audio into Accurate Text

Upload a recording or record in the browser and get back a timestamped transcript you can edit, label by speaker, and export as subtitles.

interview.mp3
00:12
00:00
Speaker 1

Welcome to Kiri TTS. Let me walk you through the recording.

00:04
Speaker 2

Sounds good — I have already uploaded the audio file.

00:09
Speaker 3

Great, the transcript is ready to export.

Export asSRTVTTJSONTXTCSV

Everything you need to transcribe audio

Speaker labels, accurate timings, an editor that keeps up, and exports that drop straight into your workflow.

Automatic speaker detection

Transcripts are split by speaker as they come back. Rename each one and give them a colour — the labels carry through to your exports.

Timestamped, editable lines

Every line keeps its own start and end time, aligned to the audio, and stays fully editable in the browser.

Subtitle-ready exports

Download as SRT, VTT, JSON, TXT or CSV. Cap the words per line so long sentences split into readable caption cues.

Trim before transcribing

Select just the section you need before you spend credits. Handles up to one hour of audio per run.

Projects and history

Group transcriptions into projects and come back to any of them later. Nothing is lost after you close the tab.

OpenAI-compatible API

Call the transcription endpoint from your own code. Existing OpenAI SDK calls work by changing the base URL.

How it works

From recording to subtitle file in three steps.

1

Upload or record

Drop in an audio file or record straight from your browser. Trim to just the part you need, up to one hour per run.

2

Get a timestamped transcript

Speakers are detected automatically and every line carries its own start and end time, aligned to the audio.

3

Edit and export

Fix any wording, rename speakers, then export as SRT, VTT, JSON, TXT or CSV.

Export formats

Take the transcript wherever it needs to go.

SRT

SubRip subtitles for video editors and players

VTT

WebVTT captions for the web and HTML5 video

JSON

Structured data with word-level timings

TXT

Plain text, with optional speakers and timestamps

CSV

Spreadsheet rows for analysis and review

Frequently Asked Questions

Does it support Khmer?

Yes. Kiri is built for Khmer first, and also transcribes English and other languages. Khmer audio returns Khmer script with correct word boundaries, not a romanised approximation.

How accurate are the timestamps?

Every transcript is passed through forced alignment after transcription, which matches each word back to the audio. That gives per-line start and end times accurate enough to use directly as subtitles.

Can it tell speakers apart?

Yes. Segments are labelled by speaker automatically. You can rename each speaker and give them a colour, and the labels carry through to your exports.

What audio files can I upload?

Common formats including MP3, WAV, M4A and OGG. You can also record directly in the browser. Each run handles up to one hour of audio, and you can trim a longer file down to the section you need.

What does transcription cost?

Transcription draws on the same monthly credits as speech generation, so there is no separate plan to buy. The free plan includes roughly 11 minutes of transcription per month, and paid plans scale from about 2.7 hours up to 55 hours.

Can I generate subtitles from it?

Yes. Export to SRT or VTT and drop the file straight into your video editor. You can also cap how many words appear per line, so long sentences are split into readable caption cues on real word timings.

Is there an API?

Yes. There is an OpenAI-compatible transcription endpoint, so existing code that calls audio.transcriptions.create works by changing the base URL. See the API docs for details.

Ready to transcribe your first recording?

Transcription uses the same credits as speech generation, so every plan includes it — the free one too.