Send real requests to the N-ATLaS models through the OpenAtlas gateway, with no SDK, no key and no sign-up. Each response is shown exactly as @openatlas/sdk returns it, next to the code that makes the same call.
OpenAtlas — PlaygroundMock backendShared demo key
Backend offline
Text generation with NCAIR1/N-ATLaS in English, Hausa, Yoruba or Igbo. The conversation stays in this page and is sent with each message.
Replies from N-ATLaS appear here.
Response
—
The same call with the SDK
Speech to text with the N-ATLaS speech recognition model for the language you pick (there is no auto-detection). Clips of up to 30 seconds.
Clip ready
TranscriptThe N-ATLaS transcript appears here.
Response
—
The same call with the SDK
Reads text aloud with a separate text-to-speech model (SoroTTS or Meta MMS-TTS), not an N-ATLaS model: it is meant for reading out what N-ATLaS has written. Clear in English; experimental in Hausa, Yoruba, Igbo and Pidgin.
Speech output unavailableThe backend's speech renderer isn't running right now.
AudioThe rendered speech appears here.
Response
—
The same call with the SDK
01
Real models, nothing swapped
Chat comes from NCAIR1/N-ATLaS and transcripts from the four N-ATLaS speech recognition models, on the same GPU backend and gateway your own key would use. A "Mock backend" badge appears if the site is ever pointed at the development mock.
02
Shared, so limited
Everyone here shares one demo key. Limits: 10 requests a minute per network, a daily quota on the key, chats of up to 10 messages and 512 tokens, clips of up to 30 seconds, speech of up to 300 characters. Your visit counts once toward the N-ATLaS 1,000-user licence cap.
03
Build on it
Copy the code next to any response, install @openatlas/sdk, and request your own key. N-ATLaS is for non-commercial use under its Terms of Use.