1{ 2 "slug": "deepgram", 3 "name": "Deepgram", 4 "description": "Speech AI platform: Nova speech-to-text (pre-recorded transcription with speaker diarization, summarization, language detection) and Aura text-to-speech (low-latency, natural voices in English and Spanish). REST API, single API key.", 5 "logo": "https://www.google.com/s2/favicons?domain=deepgram.com&sz=128", 6 "categories": ["ai", "speech", "audio", "transcription", "text-to-speech", "voice"], 7 "base_url": "https://api.deepgram.com", 8 "auth": { 9 "types": ["api_key"], 10 "headers": { 11 "Authorization": "Token {{api_key}}" 12 }, 13 "credential_fields": [ 14 { 15 "name": "api_key", 16 "label": "Deepgram API Key", 17 "description": "Create one at https://console.deepgram.com/project/default/keys. Speech calls need the normal request scopes; reporting tools need usage:read, and billing/balance/purchase tools need billing:read." 18 } 19 ] 20 }, 21 "tools": [ 22 { 23 "name": "speak", 24 "description": "Text-to-speech via Deepgram's Aura models. Converts text to natural-sounding audio at sub-second latency. Returns the generated audio as binary (mp3/wav/ogg depending on encoding). Best for conversational agents, voice notifications, and on-the-fly narration. Defaults to `aura-asteria-en` (English female); full model catalog at https://developers.deepgram.com/docs/tts-models.", 25 "method": "POST", 26 "path": "/v1/speak", 27 "timeout_ms": 60000, 28 "input_schema": { 29 "type": "object", 30 "properties": { 31 "text": { 32 "type": "string", 33 "description": "The text to convert to speech. Max 2000 characters per request — longer inputs should be split client-side." 34 }, 35 "model": { 36 "type": "string", 37 "default": "aura-asteria-en", 38 "description": "Voice model slug. Common choices: aura-asteria-en (female, conversational), aura-luna-en (female, warm), aura-stella-en (female, friendly), aura-athena-en (female, mature), aura-hera-en (female, deep), aura-orion-en (male, mature), aura-arcas-en (male, casual), aura-perseus-en (male, confident), aura-angus-en (male, Irish), aura-orpheus-en (male, narrator), aura-helios-en (male, upbeat), aura-zeus-en (male, authoritative), aura-sirio-es (Spanish male), aura-nestor-es (Spanish male). Also supports newer aura-2-* variants (e.g. aura-2-thalia-en) for higher quality." 39 }, 40 "encoding": { 41 "type": "string", 42 "enum": ["mp3", "linear16", "flac", "mulaw", "alaw", "opus", "aac"], 43 "default": "mp3", 44 "description": "Audio encoding. mp3 is the most compatible default. linear16 returns raw PCM — pair with `container=wav` to get a playable WAV file." 45 }, 46 "container": { 47 "type": "string", 48 "enum": ["none", "wav", "ogg"], 49 "description": "Audio container format. Most callers can leave this unset — mp3/linear16/flac pick sensible defaults. Required when encoding=linear16 and you want a WAV file." 50 }, 51 "sample_rate": { 52 "type": "integer", 53 "description": "Sample rate in Hz. Valid values depend on encoding: linear16/flac/mulaw/alaw accept 8000/16000/24000/32000/48000; mp3 is fixed at 22050; opus is fixed at 48000. Omit for the encoding-appropriate default." 54 }, 55 "bit_rate": { 56 "type": "integer", 57 "description": "Bit rate in bits per second for compressed encodings (mp3: 32000 or 48000, opus: 8000-32000, aac: 32000-128000). Omit for the default." 58 }, 59 "callback": { 60 "type": "string",