Skip to main content
Geko does speech in both directions, on one API key and one balance:
  • Text-to-speech — send a string and a voice name, get audio back. Speaks Kazakh today (tokay-kk-v1) in six named voices.
  • Speech-to-text — send audio, get text back. Recognises Kazakh and Russian (seta-kk-ru-v2), including a speaker switching language mid-sentence, at 8.71% WER where Whisper large-v3 scores 44.95% on the same audio.
More languages arrive on the same endpoints, so the code you write now keeps working. It’s designed to feel familiar (OpenAI / ElevenLabs style): a thin, typed SDK, a CLI, and a simple HTTP API. Pick whichever fits.

Quickstart

From an API key to audio and back to text, in under a minute.

Text-to-speech

Voices, streaming, and the quality dial.

Speech-to-text

Transcribe Kazakh/Russian audio, code-switching included.

Add it to your app

A safe server-side pattern for Next.js, Express, and beyond.

​
what you can build

Voice agents & IVR

Both ends of the loop — transcribe the caller, speak the reply, low latency.

Call & meeting transcripts

Turn recordings into searchable text, telephony audio included.

Notifications & alerts

Turn order updates, reminders, and statuses into spoken Kazakh.

Content & narration

Narrate articles, lessons, and audiobooks with named voices.

​
try it in one line

No signup code required — grab a key and run:

​
the API in a nutshell

Synthesis takes text + a voice and returns WAV (24 kHz, 16-bit PCM, mono). Transcription takes audio and returns JSON with the text. The surface is small: Base URLs. Text-to-speech: https://geko--tokay-serve-web.modal.run. Speech-to-text: https://geko--seta-serve-transcriber-api.modal.run. Two separately scaled GPU services — the SDK hides the split, so one client and one key reach both.

TypeScript SDK

npm i @gekoai/sdk → geko.tts.create(...) / geko.stt.transcribe(...).

HTTP API reference

Every endpoint, request/response shape, and error code.

Voices

Six Kazakh voices — audition and pick by ear.

Billing & credits

One balance: 0.04/1kcharacters,0.04 / 1k characters, 0.36 / audio-hour, $1.00 free to start.
New here? Grab a key at app.geko.sh — new accounts get $1.00 free — then head to the quickstart.