Talk to the model.
Listen to the reply.

Pick a voice, start the conversation and speak. Speech recognition, model and speech synthesis run in a loop, with latency for every step.

About

Kapllan Speech, live voice demo.

Kapllan Speech combines K—Listen speech recognition, K—Lite responses and K—Voice speech synthesis for English, Norwegian and Swedish. The live K—Voice checkpoint is trained by Kapllan. K—Codec is a separate development track.

Select a named English, Norwegian or Swedish K—Voice profile. Each selection binds its own authenticated speaker conditioning and the matching K—Listen language route for the complete conversation.

Speak naturally and pause briefly for a reply. The microphone remains active during replies. Speaking mutes the current reply while Kapllan verifies your words; use the X to end the conversation.