⚡ CPU-hosted · offline · developer-first

Turn speech into accurate text
with a clean developer API

Ark Transcribe converts audio to time-stamped text using OpenAI's Whisper model — running entirely on the CPU. Upload WAV, MP3, OGG or Opus, or POST to a single REST endpoint. No GPU, no third-party cloud, no audio leaving your server.

Get started — it's free View the API

Why Ark Transcribe

Built for privacy, speed, and developer ergonomics.

🧠
Whisper on the CPU

OpenAI's Whisper runs locally through whisper.cpp (ggml) — responsive transcription without a GPU or a per-minute API bill.

🌍
Automatic language detection

Leave the language on auto and Whisper identifies it for you, transcribing 99+ languages out of the box.

🔒
Your audio stays put

Decoding and transcription both happen on your server, and uploads are deleted right after processing. Nothing is sent to a third party.

⏱️
Time-stamped segments

Every result includes per-segment start/end times — perfect for subtitles, captions, and searchable transcripts.

🎧
Many formats, no codecs

WAV, MP3, OGG (Vorbis) and Opus are decoded with fully-managed libraries — cross-platform, no native ffmpeg required.

🔑
Personal API keys

Every account gets a unique key sent in a header name you choose. Regenerate it anytime from your profile.

How it works

From sign-up to your first transcript in under a minute.

1
Sign up with email

Enter your email, solve a quick CAPTCHA, and confirm the one-time code.

2
Grab your API key

A unique key is waiting on your profile, with a header name you can customize.

3
Transcribe — UI or API

Use the web transcriber or POST your audio to /api/transcribe from your app.

A REST API developers enjoy

One endpoint, a multipart upload in and JSON out, authenticated by your personal key. Pass language=auto to detect the language automatically.

  • WAV, MP3, OGG & Opus uploads up to 200 MB
  • Detected language, duration and timing returned with every response
  • Time-stamped segments for captions & subtitles
  • Live curl generator inside the transcriber
Create an account
curl -X POST https://ark-transcribe.immanuel.co/api/transcribe \ -H "X-Ark-Api-Key: ark_your_personal_key" \ -F "file=@interview.mp3" \ -F "language=auto" # → { # "text": "Hello and welcome to the show...", # "language": "en", # "durationSeconds": 92.4, # "processingSeconds": 11.8, # "segments": [ # { "start": 0.0, "end": 3.2, "text": "Hello and welcome to the show" } # ] # }