How-to

Audio and background

Background sound, noise filtering, and keyword filtering for cleaner audio output.

Overview

Breeze Buddy provides three audio-related configuration areas: background ambient sound, noise-enhancement filtering on incoming audio, and keyword suppression on user transcriptions while the bot is speaking.

Background sound

Mix ambient audio into the call to make the interaction feel more natural — for example, simulating a busy office environment.

Field	Type	Default	Description
`enable_background_sound`	`bool`	`false`	Enable or disable ambient background audio.
`background_sound_file`	`BackgroundSoundFile`	`null`	Which ambient file to play. Currently available: `office-ambience`.
`background_sound_volume`	`float`	`2.0`	Gain multiplier for the background sound level.

json

{
  "configurations": {
    "enable_background_sound": true,
    "background_sound_file": "office-ambience",
    "background_sound_volume": 2.0
  }
}

Volume tuning

The default volume of 2.0 works well for most scenarios. Lower it to 1.0–1.5 if callers report difficulty hearing the bot’s voice over the ambient sound.

Noise filter

The noise filter applies audio enhancement to the incoming audio stream, reducing background noise and improving STT accuracy.

Field	Type	Default	Description
`enable`	`bool`	`false`	Enable or disable the noise filter.
`type`	`NoiseFilterType`	`aic`	Noise filter algorithm. Currently available: `aic`.

NoiseFilterType

Value	Description
`aic`	AIC-based noise enhancement. Cleans up background noise, echo, and ambient interference for clearer audio input.

json

{
  "configurations": {
    "noise_filter": {
      "enable": true,
      "type": "aic"
    }
  }
}

When to use

Enable AIC noise filtering when callers are in noisy environments (vehicles, outdoors, crowded areas). It adds minimal latency and significantly improves STT accuracy.

Keyword filter

The keyword filter drops incoming user transcriptions that match specific words or phrases while the bot is actively speaking. It’s a surgical tool for stopping short filler responses (“hmm”, “okay”, “hello”) from interrupting the bot mid-sentence. It does not affect the bot’s own speech output.

See the full keyword filter page for the detailed behaviour, including how it interacts with min_words and disabled_discard.

Field	Type	Default	Description
`enabled`	`bool`	`false`	Master switch.
`keywords`	`List[str]`	`[]`	Words or phrases to suppress (case-insensitive, whitespace-trimmed).
`match_type`	`KeywordMatchType`	`exact`	Matching strategy: `exact` or `includes`.

KeywordMatchType

Value	Description	Example
`exact`	The entire transcription must equal the keyword.	`"okay"` matches `"okay"` but not `"okay sure"`.
`includes`	Substring match — the transcription contains the keyword.	`"okay"` matches `"okay"` and `"okay sure"`.

json

{
  "configurations": {
    "keyword_filter": {
      "enabled": true,
      "keywords": ["hmm", "okay", "hello"],
      "match_type": "exact"
    }
  }
}

Use exact matching for short words

Prefer exact for short acknowledgement words. includes aggressively drops any transcription containing the keyword — “okay, what’s my balance” would be silently dropped.

Full audio configuration example

json

{
  "configurations": {
    "enable_background_sound": true,
    "background_sound_file": "office-ambience",
    "background_sound_volume": 1.5,
    "noise_filter": {
      "enable": true,
      "type": "aic"
    },
    "keyword_filter": {
      "enabled": true,
      "keywords": ["um", "uh", "hmm"],
      "match_type": "exact"
    }
  }
}

Was this helpful?

Edit on GitHub