CLI

@gekoai/sdk ships a CLI, so you can use both directions of the API without writing code.

# run without installing
npx @gekoai/sdk say "Сәлеметсіз бе!" --voice Aigerim -o hello.wav
npx @gekoai/sdk transcribe hello.wav

# or install globally → `gekoai`
npm install -g @gekoai/sdk
gekoai say "Сәлем" -o hello.wav

Set your key once:

export GEKO_API_KEY=sk-tokay-...

Commands

`say <text>`

Synthesize text to a .wav file.

gekoai say "Тапсырыс нөмірі 152 дайын." --voice Arman --nfe 32 -o order.wav

Option	Description
`--voice <name>`	Voice (default: the model’s default, `Aigerim`).
`--model <id>`	Model id (default: `tokay-kk-v1`).
`--nfe <n>`	Diffusion steps: `16` fast, `32` quality.
`--speed <n>`	Speed multiplier (0.5–2).
`--no-normalize`	Disable number/currency/date normalization.
`--out, -o <file>`	Output path (default: `speech.wav`).

`transcribe <file>`

Transcribe a local audio file, or a public URL with --url.

gekoai transcribe call.wav
gekoai transcribe --url https://example.com/call.mp3 --json

Option	Description
`--url <url>`	Transcribe a public audio URL instead of a local file.
`--stt-base-url <url>`	Override the transcription API base URL (or set `GEKO_STT_BASE_URL`).
`--json`	Print the full response — timings, `audio_seconds`, `credits_charged`.

Pass a file or --url, not both. Accepts any container ffmpeg reads (wav, mp3, m4a, flac, ogg, opus), up to 2 hours per request.

The transcript goes to stdout and the cost line to stderr, so redirecting captures only the text:

gekoai transcribe call.wav > transcript.txt
#   2.84s audio · 8 credits · 18.5× realtime   ← stderr, still on your terminal

`voices`

List voices for a model (no key required).

gekoai voices
gekoai voices --model tokay-kk-v1 --json

`models`

List available text-to-speech models (no key required).

gekoai models
gekoai models --json

Global flags

Flag	Description
`--api-key <key>`	Override `$GEKO_API_KEY`.
`--base-url <url>`	Override the text-to-speech base URL (or set `GEKO_BASE_URL`).
`--json`	Machine-readable output.
`--help, -h`	Show help.
`--version, -v`	Print the SDK version.

Exit codes

Code	Meaning
`0`	Success.
`1`	API error (the server responded with an error, or the request failed).
`2`	Usage error (missing text/audio, missing key, unknown command).

Scripting examples

Generate a clip per line of a file:

i=0
while IFS= read -r line; do
  gekoai say "$line" --voice Aigerim -o "clip_$i.wav"
  i=$((i + 1))
done < lines.txt

Transcribe a directory of recordings into one transcript each:

for f in recordings/*.mp3; do
  gekoai transcribe "$f" > "${f%.mp3}.txt"
done

Next steps

TypeScript SDK

The programmatic client behind the same CLI.

Authentication

Create a key and set GEKO_API_KEY.

Pick a voice

Browse the catalog and audition each voice.

Errors & retries

What each exit code maps to under the hood.

get started

text-to-speech

speech-to-text

guides

sdk

reference

Commands

`say <text>`

`transcribe <file>`

`voices`

`models`

Global flags

Exit codes

Scripting examples

Next steps

TypeScript SDK

Authentication

Pick a voice

Errors & retries

​Commands

​say <text>

​transcribe <file>

​voices

​models

​Global flags

​Exit codes

​Scripting examples

​Next steps

TypeScript SDK

Authentication

Pick a voice

Errors & retries

Commands

`say <text>`

`transcribe <file>`

`voices`

`models`

Global flags

Exit codes

Scripting examples

Next steps