Speech drivers

This page documents the speech synthesis plugin in fp_perception_driver_speech.

OpenAISpeechDriver

Class: fp_perception::OpenAISpeechDriver (REST-based TTS)

What it does

• Accepts fp_perception::text_data via synthesize().
• Uses RestBase::call_tts() to request PCM audio from the configured REST endpoint.
• Splits long text into smaller word-bounded chunks, synthesizes those chunks with bounded parallelism, then concatenates the PCM so playback stays in the original text order.
• Returns the synthesized audio as fp_perception::audio_data.

Parameters

• driver.speech.OpenAISpeechDriver.name (string)
• driver.speech.OpenAISpeechDriver.model (string)
• driver.speech.OpenAISpeechDriver.test_text (string)
• driver.speech.OpenAISpeechDriver.test_file_path (string)
• driver.speech.OpenAISpeechDriver.voice (string)
• driver.speech.OpenAISpeechDriver.instructions (string)
• driver.speech.OpenAISpeechDriver.chunking_enabled (bool): enable long-text chunking before TTS requests.
• driver.speech.OpenAISpeechDriver.chunk_max_words (int): maximum number of words sent in a single TTS request.
• driver.speech.OpenAISpeechDriver.chunk_parallel_requests (int): maximum number of chunk requests to synthesize concurrently.

REST base parameters:

• driver.speech.OpenAISpeechDriver.rest.uri
• driver.speech.OpenAISpeechDriver.rest.method
• driver.speech.OpenAISpeechDriver.rest.ssl_verify
• driver.speech.OpenAISpeechDriver.rest.auth_type

Environment:

• Requires OPENAI_API_KEY to be set.

Usage with the server

• The server offers fp_perception_msgs/srv/PerceptionSpeech.
• If use_device_audio=true, the server routes synthesized audio to the speaker driver.
• Otherwise, it returns the audio in the service response.
• If use_diagnostics=true, the driver publishes request health on /diagnostics via diagnostic_updater.