> ## Documentation Index
> Fetch the complete documentation index at: https://docs.pinkfish.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# audio-processing
> Text-to-speech generation and audio/video transcription
**Server path:** `/audio-processing` | **Type:** Embedded | **PCID required:** No
## Tools
| Tool | Description |
| ------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| [`audio-processing_text_to_speech`](#audio-processing_text_to_speech) | Convert text to speech using minimax/speech-02-turbo model. Generates high-quality audio from text input with configurable voice settings. Returns a URL to the generated audio file. |
| [`audio-processing_transcribe_audio_or_video`](#audio-processing_transcribe_audio_or_video) | Transcribe audio or video files to text using Deepgram AI. Returns transcription text plus optional word-level / utterance / paragraph / speaker timestamps via the includeDetailedResults flag. Supports diarization, keyword boosting (keyterm for Nova-3), search, replace, multichannel, filler words, and more. |
***
## audio-processing\_text\_to\_speech
Convert text to speech using minimax/speech-02-turbo model. Generates high-quality audio from text input with configurable voice settings. Returns a URL to the generated audio file.
**Parameters:**
| Parameter | Type | Required | Default | Description |
| ---------------- | ------ | -------- | -------------- | -------------------------------------------------------------------------------------------------------------------- |
| `text` | string | Yes | — | Text to convert to speech. Maximum 5000 characters. Use \<#x#> between words to control pause duration (0.01-99.99s) |
| `voice_id` | string | No | `"Wise_Woman"` | Voice ID for text-to-speech generation |
| `pitch` | number | No | — | Speech pitch (-12 to 12, default: 0) |
| `speed` | number | No | — | Speech speed multiplier (0.5 to 2, default: 1) |
| `volume` | number | No | — | Speech volume level (0 to 10, default: 1) |
| `emotion` | string | No | — | Emotion to apply to speech (default: auto) |
| `sample_rate` | number | No | — | Audio sample rate (default: 32000) |
| `language_boost` | string | No | — | Language enhancement for better pronunciation |
```json theme={null}
{
"type": "object",
"properties": {
"text": {
"type": "string",
"description": "Text to convert to speech. Maximum 5000 characters. Use <#x#> between words to control pause duration (0.01-99.99s)"
},
"voice_id": {
"type": "string",
"enum": [
"Wise_Woman",
"Friendly_Person",
"Inspirational_girl",
"Deep_Voice_Man",
"Calm_Woman",
"Casual_Guy",
"Lively_Girl",
"Patient_Man",
"Young_Knight",
"Determined_Man",
"Lovely_Girl",
"Decent_Boy",
"Imposing_Manner",
"Elegant_Man",
"Abbess",
"Sweet_Girl_2",
"Exuberant_Girl"
],
"default": "Wise_Woman",
"description": "Voice ID for text-to-speech generation"
},
"pitch": {
"type": "number",
"default": 0,
"description": "Speech pitch (-12 to 12, default: 0)"
},
"speed": {
"type": "number",
"default": 1,
"description": "Speech speed multiplier (0.5 to 2, default: 1)"
},
"volume": {
"type": "number",
"default": 1,
"description": "Speech volume level (0 to 10, default: 1)"
},
"emotion": {
"type": "string",
"enum": [
"auto",
"neutral",
"happy",
"sad",
"angry",
"fearful",
"disgusted",
"surprised"
],
"default": "auto",
"description": "Emotion to apply to speech (default: auto)"
},
"sample_rate": {
"type": "number",
"default": 32000,
"description": "Audio sample rate (default: 32000)"
},
"language_boost": {
"type": "string",
"enum": [
"None",
"Automatic",
"Chinese",
"Chinese,Yue",
"English",
"Arabic",
"Russian",
"Spanish",
"French",
"Portuguese",
"German",
"Turkish",
"Dutch",
"Ukrainian",
"Vietnamese",
"Indonesian",
"Japanese",
"Italian",
"Korean",
"Thai",
"Polish",
"Romanian",
"Greek",
"Czech",
"Finnish",
"Hindi"
],
"default": "None",
"description": "Language enhancement for better pronunciation"
}
},
"required": [
"text"
]
}
```
***
## audio-processing\_transcribe\_audio\_or\_video
Transcribe audio or video files to text using Deepgram AI. Returns transcription text plus optional word-level / utterance / paragraph / speaker timestamps via the includeDetailedResults flag. Supports diarization, keyword boosting (keyterm for Nova-3), search, replace, multichannel, filler words, and more.
**Parameters:**
| Parameter | Type | Required | Default | Description |
| ------------------------- | --------- | -------- | ------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `fileUrl` | string | Yes | — | URL of the audio or video file to transcribe |
| `model` | string | No | — | Deepgram model. nova-3 is the latest (required for keyterm). Whisper tiers available. |
| `languageCode` | string | No | — | Language code (e.g., "en", "es", "fr"). Omit for auto-detect. |
| `detectLanguage` | boolean | No | — | Explicitly request language auto-detection (most models detect by default when languageCode is omitted) |
| `smartFormat` | boolean | No | — | Smart formatting of numbers, dates, times, etc. Default true. |
| `numerals` | boolean | No | — | Convert number words to numerals |
| `profanityFilter` | boolean | No | — | Filter profanity |
| `dictation` | boolean | No | — | Interpret spoken punctuation commands |
| `measurements` | boolean | No | — | Normalize measurement abbreviations |
| `fillerWords` | boolean | No | — | Include "um" / "uh" / etc. in transcript |
| `enableDiarization` | boolean | No | — | Identify different speakers. Auto-enables utterances with speaker-attributed timestamps. |
| `diarizationSpeakerCount` | number | No | — | Hint for expected number of speakers |
| `diarizeVersion` | string | No | — | Pin diarization algorithm version |
| `enableUtterances` | boolean | No | — | Group output into utterances with start/end timestamps per phrase/speaker. Auto-enabled when diarization is on. Compact — does not include per-word timestamps. |
| `uttSplit` | number | No | — | Silence threshold in seconds for splitting utterances (default 0.8) |
| `enableParagraphs` | boolean | No | — | Format output into paragraphs with timestamps |
| `multichannel` | boolean | No | — | Transcribe each audio channel separately. Critical for stereo call recordings. |
| `alternatives` | number | No | — | Return N transcription candidates per channel |
| `keywords` | string\[] | No | — | Keyword boosting. Each entry may be "word" or "word:intensifier". Pre-Nova-3 models. |
| `keyterm` | string\[] | No | — | Nova-3 keyterm boosting (superior to keywords; supports multi-word phrases). English-only. |
| `search` | string\[] | No | — | Search for phrases; hits returned with timestamps |
| `replace` | string\[] | No | — | Find-and-replace in transcript. Format: "find:replace". |
| `enableSummary` | boolean | No | — | Generate a short text summary of the audio. Requires 50+ words of audio — shorter inputs return the original text. Returns a summary.short field. |
| `enableTopics` | boolean | No | — | Detect topics discussed in the audio. Returns topic segments with labels and confidence scores (0-1). Example topics: "healthcare", "data collection", "budget". English only. |
| `customTopics` | string\[] | No | — | Provide up to 100 custom topics to detect. By default uses extended mode — returns your topics plus auto-detected ones. Requires enableTopics. Example: \["sales", "support", "billing", "technical issue"] |
| `enableIntents` | boolean | No | — | Detect speaker intents — what speakers are trying to do. Returns intent segments with verb-form labels and confidence scores (0-1). Example intents: "schedule a meeting", "request a refund". English only. |
| `customIntents` | string\[] | No | — | Provide up to 100 custom intents to detect. By default uses extended mode — returns your intents plus auto-detected ones. Requires enableIntents. Example: \["purchase", "cancel subscription", "get status update", "file complaint"] |
| `enableSentiment` | boolean | No | — | Analyze sentiment per segment — tags each part as positive, negative, or neutral with a confidence score (0-1). Also provides an average sentiment for the entire transcript. Useful for call center analysis. |
| `redact` | string\[] | No | — | Redaction options. Common: pci, pii, phi, numbers, ssn. Also supports specific entity types. |
| `mipOptOut` | boolean | No | — | Opt out of Deepgram model improvement program |
| `tag` | string\[] | No | — | Labels for Deepgram console analytics filtering |
| `extra` | string\[] | No | — | Arbitrary "key:value" metadata passthrough |
| `outputMode` | string | No | — | Output mode. "transcription" (default): text only (\~5KB). "transcription\_with\_timestamps": includes utterance-level timestamps with start/end per phrase/speaker (\~15KB). Use timestamps mode when you need to know when things were said. |
| `includeDetailedResults` | boolean | No | — | Include the full raw Deepgram response with word-level timestamps. Warning: very large output (200KB+ per minute of audio). Only use when you need per-word timing data. |
```json theme={null}
{
"type": "object",
"properties": {
"fileUrl": {
"type": "string",
"description": "URL of the audio or video file to transcribe"
},
"model": {
"type": "string",
"enum": [
"nova-3",
"nova-3-general",
"nova-3-medical",
"nova-2",
"nova-2-general",
"nova-2-meeting",
"nova-2-phonecall",
"nova-2-voicemail",
"nova-2-finance",
"nova-2-conversationalai",
"nova-2-video",
"nova-2-medical",
"nova-2-drivethru",
"nova-2-automotive",
"nova",
"enhanced",
"base",
"whisper-tiny",
"whisper-base",
"whisper-small",
"whisper-medium",
"whisper-large"
],
"default": "nova-3",
"description": "Deepgram model. nova-3 is the latest (required for keyterm). Whisper tiers available."
},
"languageCode": {
"type": "string",
"description": "Language code (e.g., \"en\", \"es\", \"fr\"). Omit for auto-detect."
},
"detectLanguage": {
"type": "boolean",
"description": "Explicitly request language auto-detection (most models detect by default when languageCode is omitted)"
},
"smartFormat": {
"type": "boolean",
"description": "Smart formatting of numbers, dates, times, etc. Default true."
},
"numerals": {
"type": "boolean",
"description": "Convert number words to numerals"
},
"profanityFilter": {
"type": "boolean",
"description": "Filter profanity"
},
"dictation": {
"type": "boolean",
"description": "Interpret spoken punctuation commands"
},
"measurements": {
"type": "boolean",
"description": "Normalize measurement abbreviations"
},
"fillerWords": {
"type": "boolean",
"description": "Include \"um\" / \"uh\" / etc. in transcript"
},
"enableDiarization": {
"type": "boolean",
"default": false,
"description": "Identify different speakers. Auto-enables utterances with speaker-attributed timestamps."
},
"diarizationSpeakerCount": {
"type": "number",
"description": "Hint for expected number of speakers"
},
"diarizeVersion": {
"type": "string",
"description": "Pin diarization algorithm version"
},
"enableUtterances": {
"type": "boolean",
"description": "Group output into utterances with start/end timestamps per phrase/speaker. Auto-enabled when diarization is on. Compact — does not include per-word timestamps."
},
"uttSplit": {
"type": "number",
"description": "Silence threshold in seconds for splitting utterances (default 0.8)"
},
"enableParagraphs": {
"type": "boolean",
"default": false,
"description": "Format output into paragraphs with timestamps"
},
"multichannel": {
"type": "boolean",
"description": "Transcribe each audio channel separately. Critical for stereo call recordings."
},
"alternatives": {
"type": "number",
"description": "Return N transcription candidates per channel"
},
"keywords": {
"type": "array",
"items": {
"type": "string"
},
"description": "Keyword boosting. Each entry may be \"word\" or \"word:intensifier\". Pre-Nova-3 models."
},
"keyterm": {
"type": "array",
"items": {
"type": "string"
},
"description": "Nova-3 keyterm boosting (superior to keywords; supports multi-word phrases). English-only."
},
"search": {
"type": "array",
"items": {
"type": "string"
},
"description": "Search for phrases; hits returned with timestamps"
},
"replace": {
"type": "array",
"items": {
"type": "string"
},
"description": "Find-and-replace in transcript. Format: \"find:replace\"."
},
"enableSummary": {
"type": "boolean",
"default": false,
"description": "Generate a short text summary of the audio. Requires 50+ words of audio — shorter inputs return the original text. Returns a summary.short field."
},
"enableTopics": {
"type": "boolean",
"default": false,
"description": "Detect topics discussed in the audio. Returns topic segments with labels and confidence scores (0-1). Example topics: \"healthcare\", \"data collection\", \"budget\". English only."
},
"customTopics": {
"type": "array",
"items": {
"type": "string"
},
"description": "Provide up to 100 custom topics to detect. By default uses extended mode — returns your topics plus auto-detected ones. Requires enableTopics. Example: [\"sales\", \"support\", \"billing\", \"technical issue\"]"
},
"enableIntents": {
"type": "boolean",
"description": "Detect speaker intents — what speakers are trying to do. Returns intent segments with verb-form labels and confidence scores (0-1). Example intents: \"schedule a meeting\", \"request a refund\". English only."
},
"customIntents": {
"type": "array",
"items": {
"type": "string"
},
"description": "Provide up to 100 custom intents to detect. By default uses extended mode — returns your intents plus auto-detected ones. Requires enableIntents. Example: [\"purchase\", \"cancel subscription\", \"get status update\", \"file complaint\"]"
},
"enableSentiment": {
"type": "boolean",
"default": false,
"description": "Analyze sentiment per segment — tags each part as positive, negative, or neutral with a confidence score (0-1). Also provides an average sentiment for the entire transcript. Useful for call center analysis."
},
"redact": {
"type": "array",
"items": {
"type": "string",
"enum": [
"pci",
"pii",
"phi",
"numbers",
"ssn",
"aggressive_numbers",
"credit_card",
"credit_card_expiration",
"cvv",
"email_address",
"phone_number",
"account_number",
"age",
"date",
"date_interval",
"dob",
"driver_license",
"healthcare_number",
"ip_address",
"location",
"location_address",
"location_zip",
"location_coordinate",
"money",
"numerical_pii",
"passport_number",
"password",
"time",
"vehicle_id",
"statistics",
"bank_account",
"routing_number"
]
},
"description": "Redaction options. Common: pci, pii, phi, numbers, ssn. Also supports specific entity types."
},
"mipOptOut": {
"type": "boolean",
"description": "Opt out of Deepgram model improvement program"
},
"tag": {
"type": "array",
"items": {
"type": "string"
},
"description": "Labels for Deepgram console analytics filtering"
},
"extra": {
"type": "array",
"items": {
"type": "string"
},
"description": "Arbitrary \"key:value\" metadata passthrough"
},
"outputMode": {
"type": "string",
"enum": [
"transcription",
"transcription_with_timestamps"
],
"default": "transcription",
"description": "Output mode. \"transcription\" (default): text only (~5KB). \"transcription_with_timestamps\": includes utterance-level timestamps with start/end per phrase/speaker (~15KB). Use timestamps mode when you need to know when things were said."
},
"includeDetailedResults": {
"type": "boolean",
"description": "Include the full raw Deepgram response with word-level timestamps. Warning: very large output (200KB+ per minute of audio). Only use when you need per-word timing data."
}
},
"required": [
"fileUrl"
]
}
```