Repository navigation
Support OpenAI's v1/realtime transcription API #588
Description
Describe the feature you'd like
Implement the Websockets based v1/realtime API defined by OpenAI.
See https://developers.openai.com/api/docs/guides/realtime-transcription
Describe alternatives you've considered
The already implemented /v1/audio/speech API works great, but latency could be reduced and output streamed if the realtime API were to be implemented.
Additional context
v1/realtimewas implemented in llama-swap with internal/server: add /v1/realtime proxying and a Realtime playground mostlygeek/llama-swap#1299v1/realtimewas implemented in llama.cpp with server : add realtime conversation and transcription on /v1/realtime ggml-org/llama.cpp#31269
Supported by clients such as OpenWebUI and Home Assistant via the Wyoming adapter at https://github.com/roryeckel/wyoming_openai
Activity
remsky commentedon Nov 5, 2026
Hey @candrews, it was added in the master branch recently actually alongside Cartesia/Elevenlabs websocket support; next pre-built release will be out shortly with it. Feel free to re-open if any issues come up with it.
#584