Recommended settings

Use the speech recognition and translation settings below to get started with the API. For detailed description and fine-tuning please refer to [translation settings breakdown section][2] or contact Palabra.ai support. Below you can find default, speed, and quality settings preset.

Default preset

Optimal settings to start with.

{
  "message_type": "set_task",
  "data": {
    "input_stream": {
      "content_type": "audio",
      "source": {
        "type": "livekit"
      }
    },
    "output_stream": {
      "content_type": "audio",
      "target": {
        "type": "livekit"
      }
    },
    "pipeline": {
      "preprocessing": {
        "enable_vad": true,
        "vad_threshold": 0.5,
        "vad_left_padding": 1,
        "vad_right_padding": 1,
        "pre_vad_denoise": false,
        "pre_vad_dsp": true
      },
      "transcription": {
        "source_language": "en",
        "detectable_languages": [],
        "asr_model": "auto",
        "denoise": "none",
        "allow_hotwords_glossaries": true,
        "suppress_numeral_tokens": false,
        "diarize_speakers": false,
        "priority": "normal",
        "min_alignment_score": 0.15,
        "max_alignment_cer": 0.8,
        "segment_confirmation_silence_threshold": 0.7,
        "only_confirm_by_silence": false,
        "batched_inference": false,
        "force_detect_language": false,
        "sentence_splitter": {
          "enabled": true,
          "splitter_model": "auto"
        },
        "verification": {
          "verification_model": "auto",
          "allow_verification_glossaries": true,
          "auto_transcription_correction": false,
          "transcription_correction_style": null
        }
      },
      "translations": [
        {
          "target_language": "es",
          "allowed_source_languages": [],
          "translation_model": "auto",
          "allow_translation_glossaries": true,
          "style": null,
          "speech_generation": {
            "tts_model": "auto",
            "voice_cloning": false,
            "voice_id": "default_low",
            "voice_timbre_detection": {
              "enabled": false,
              "high_timbre_voices": [
                "default_high"
              ],
              "low_timbre_voices": [
                "default_low"
              ]
            },
            "denoise_voice_samples": false,
            "speech_tempo_auto": true,
            "speech_tempo_timings_factor": 0,
            "speech_tempo_adjustment_factor": 0.77
          }
        }
      ],
      "translation_queue_configs": {
        "global": {
          "desired_queue_level_ms": 8000,
          "max_queue_level_ms": 24000
        }
      },
      "allowed_message_types": [
        "partial_transcription",
        "validated_transcription",
        "translated_transcription"
      ]
    }
  }
}

Speed preset

Minimum latency preset.

{
  "message_type": "set_task",
  "data": {
    "input_stream": {
      "content_type": "audio",
      "source": {
        "type": "livekit"
      }
    },
    "output_stream": {
      "content_type": "audio",
      "target": {
        "type": "livekit"
      }
    },
    "pipeline": {
      "preprocessing": {
        "enable_vad": true,
        "vad_threshold": 0.5,
        "vad_left_padding": 1,
        "vad_right_padding": 1,
        "pre_vad_denoise": false,
        "pre_vad_dsp": true
      },
      "transcription": {
        "source_language": "en",
        "detectable_languages": [],
        "asr_model": "alpha_turbo", // changed from "auto" in the default preset
        "denoise": "none", // must be "none", affects performance
        "allow_hotwords_glossaries": true,
        "suppress_numeral_tokens": false,
        "diarize_speakers": false, // must be false, affects performance
        "priority": "normal",
        "min_alignment_score": 0.2, // changed from 0.15 in the default preset
        "max_alignment_cer": 0.8,
        "segment_confirmation_silence_threshold": 0.55, // changed from 0.7 in the default preset
        "only_confirm_by_silence": false,
        "batched_inference": true, // changed from "false" in the default preset
        "force_detect_language": false, // must be false, affects performance
        "sentence_splitter": {
          "enabled": true,
          "splitter_model": "auto"
        },
        "verification": {
          "verification_model": "auto",
          "allow_verification_glossaries": true,
          "auto_transcription_correction": false,
          "transcription_correction_style": null
        }
      },
      "translations": [
        {
          "target_language": "es",
          "allowed_source_languages": [],
          "translation_model": "auto",
          "allow_translation_glossaries": true,
          "style": null,
          "speech_generation": {
            "tts_model": "auto",
            "voice_cloning": false, // must be "false", affects performance
            "voice_id": "default_low",
            "voice_timbre_detection": {
              "enabled": false, // must be "false", affects performance
              "high_timbre_voices": [
                "default_high"
              ],
              "low_timbre_voices": [
                "default_low"
              ]
            },
            "denoise_voice_samples": false,
            "speech_tempo_auto": true,
            "speech_tempo_timings_factor": 0,
            "speech_tempo_adjustment_factor": 0.77
          }
        }
      ],
      "translation_queue_configs": {
        "global": {
          "desired_queue_level_ms": 8000,
          "max_queue_level_ms": 24000
        }
      },
      "allowed_message_types": [
        "partial_transcription",
        "validated_transcription",
        "translated_transcription"
      ]
    }
  }
}

Quality preset

Maximum quality preset.

{
  "message_type": "set_task",
  "data": {
    "input_stream": {
      "content_type": "audio",
      "source": {
        "type": "livekit"
      }
    },
    "output_stream": {
      "content_type": "audio",
      "target": {
        "type": "livekit"
      }
    },
    "pipeline": {
      "preprocessing": {
        "enable_vad": true,
        "vad_threshold": 0.5,
        "vad_left_padding": 1,
        "vad_right_padding": 1,
        "pre_vad_denoise": false,
        "pre_vad_dsp": true
      },
      "transcription": {
        "source_language": "en",
        "detectable_languages": [],
        "asr_model": "auto",
        "denoise": "none",
        "allow_hotwords_glossaries": true,
        "suppress_numeral_tokens": false,
        "diarize_speakers": false,
        "priority": "quality", // changed from "normal" in the default preset
        "min_alignment_score": 0.15,
        "max_alignment_cer": 0.8,
        "segment_confirmation_silence_threshold": 0.85, // changed from 0.7 in the default preset
        "only_confirm_by_silence": false,
        "batched_inference": false,
        "force_detect_language": false,
        "sentence_splitter": {
          "enabled": true,
          "splitter_model": "auto"
        },
        "verification": {
          "verification_model": "auto",
          "allow_verification_glossaries": true,
          "auto_transcription_correction": false,
          "transcription_correction_style": null
        }
      },
      "translations": [
        {
          "target_language": "es",
          "allowed_source_languages": [],
          "translation_model": "auto",
          "allow_translation_glossaries": true,
          "style": null,
          "speech_generation": {
            "tts_model": "auto",
            "voice_cloning": false,
            "voice_id": "default_low",
            "voice_timbre_detection": {
              "enabled": false,
              "high_timbre_voices": [
                "default_high"
              ],
              "low_timbre_voices": [
                "default_low"
              ]
            },
            "denoise_voice_samples": false,
            "speech_tempo_auto": true,
            "speech_tempo_timings_factor": 0,
            "speech_tempo_adjustment_factor": 0.77
          }
        }
      ],
      "translation_queue_configs": {
        "global": {
          "desired_queue_level_ms": 8000,
          "max_queue_level_ms": 24000
        }
      },
      "allowed_message_types": [
        "partial_transcription",
        "validated_transcription",
        "translated_transcription"
      ]
    }
  }
}