# Core.Today AI API — Speech-to-Text (STT) Models

> Speech recognition models (audio in, text out — Whisper, Meta Muse Voice Transcribe) available through the Core.Today unified API. File transcription is POST /v1/predictions with an audio URL; realtime streaming is wss://api.core.today/v1/realtime/asr; OpenAI SDK clients can call whisper-1 at https://api.core.today/llm/openai/v1/audio/transcriptions with a multipart file. For text in, audio out see /llms/audio-models.txt.

Source: https://console.core.today/docs/speech-to-text
This file: https://console.core.today/llms/speech-to-text.txt
All models: https://console.core.today/llms.txt

## API Usage

Base URL: https://api.core.today/v1
Auth header: X-API-Key: YOUR_API_KEY

1. POST /predictions with {"model": "<model id>", "input": {...}} -> returns job_id
2. GET /predictions/{job_id} -> poll until status is "completed" or "failed"
3. result contains output URL(s); add "is_public": true for a permanent public URL

## Models (7)

### GPT Transcribe (OpenAI API) — `openai/gpt-transcribe`

OpenAI's gpt-transcribe speech recognition, the successor to whisper-1 - plain transcript text with detected languages, optional language hints and keyword spelling. No timestamps. Billed per second of audio (0.174375 credits/s, about 10.5 credits per minute), up to 1 hour / 25MB per request.

- Provider: OpenAI
- Credits: 11 (per minute of audio (0.174375/s, settled on measured length))
- Speed: fast / Quality: high
- Required params: audio (string)
- Optional params: response_format, languages, keywords, prompt
- Example input:
```json
{
  "model": "openai/gpt-transcribe",
  "input": {
    "audio": "https://replicate.delivery/mgxm/e5159b1b-508a-4be4-b892-e1eb47850bdc/OSR_uk_000_0050_8k.wav",
    "languages": [
      "en"
    ]
  }
}
```
- Full docs: https://console.core.today/docs/models/openai/gpt-transcribe

### Incredibly Fast Whisper — `vaibhavs10/incredibly-fast-whisper`

Whisper large-v3 optimized for speed (38M+ runs) — transcribes roughly 150 minutes of audio in under 100 seconds using batched inference. Chunk-level or word-level timestamps.

- Provider: Whisper
- Credits: 10 (per run)
- Speed: fast / Quality: ultra
- Required params: audio (string)
- Optional params: task=transcribe, language=None, batch_size=24, timestamp=chunk, diarise_audio=false, hf_token
- Example input:
```json
{
  "model": "vaibhavs10/incredibly-fast-whisper",
  "input": {
    "audio": "https://example.com/podcast-episode.mp3",
    "task": "transcribe",
    "timestamp": "word"
  }
}
```
- Full docs: https://console.core.today/docs/models/vaibhavs10/incredibly-fast-whisper

### Muse Voice Transcribe 1.0 — `meta/muse-voice-transcribe-1.0`

Meta's Muse Voice Transcribe speech-to-text over the Meta developer API - 25 languages with code-switching (Korean included), speaker diarization, endpointing, keyword and language biasing. Billed per second of audio (0.11625 credits/s, about 7 credits per minute), up to 10 minutes per request.

- Provider: Meta
- Credits: 7 (per minute of audio (0.11625/s, settled on measured length))
- Speed: fast / Quality: high
- Required params: audio (string)
- Optional params: mode=PUSH_TO_TALK, keywords, language_bias
- Example input:
```json
{
  "model": "meta/muse-voice-transcribe-1.0",
  "input": {
    "audio": "https://replicate.delivery/mgxm/e5159b1b-508a-4be4-b892-e1eb47850bdc/OSR_uk_000_0050_8k.wav",
    "mode": "PUSH_TO_TALK",
    "language_bias": [
      "English"
    ]
  }
}
```
- Full docs: https://console.core.today/docs/models/meta/muse-voice-transcribe-1.0

### Muse Voice Transcribe 1.0 (Realtime) — `meta/muse-voice-transcribe-1.0/realtime`

Meta's Muse Voice Transcribe over a WebSocket - stream 16-bit PCM (24 kHz or 16 kHz mono) and receive partial and final transcripts while you speak, with endpointing and diarization. Billed per second of audio Meta processed (0.11625 credits/s, 35 credits per 5-minute block), sessions up to 20 minutes.

- Provider: Meta
- Credits: 35 (per 5-minute block (0.11625 credits/s))
- Speed: fast / Quality: high
- Required params: model (string)
- Optional params: audioEncoding=PCM_24KHZ, mode=PUSH_TO_TALK, partialMode=CUMULATIVE, emitAudioProgress=false, keywords, languageBias, customerId, apiKey
- Example input:
```json
{
  "model": "meta/muse-voice-transcribe-1.0/realtime",
  "input": {
    "model": "meta/muse-voice-transcribe-1.0/realtime",
    "audioEncoding": "PCM_24KHZ",
    "mode": "ENDPOINTING",
    "partialMode": "CUMULATIVE",
    "emitAudioProgress": false,
    "keywords": [
      "Core.Today"
    ],
    "languageBias": [
      "Korean"
    ],
    "customerId": "member-123"
  }
}
```
- Full docs: https://console.core.today/docs/models/meta/muse-voice-transcribe-1.0/realtime

### Whisper-1 (OpenAI API) — `openai/whisper-1`

OpenAI's hosted whisper-1 speech recognition with the OpenAI API parameters - json or verbose_json output, segment and word timestamps (timestamp_granularities), language hint and prompt. Billed per second of audio (0.2325 credits/s, about 14 credits per minute), up to 1 hour / 25MB per request.

- Provider: OpenAI
- Credits: 14 (per minute of audio (0.2325/s, settled on measured length))
- Speed: fast / Quality: high
- Status: DEPRECATED (may be removed)
- Required params: audio (string)
- Optional params: response_format=json, timestamp_granularities, language, prompt, temperature=0
- Example input:
```json
{
  "model": "openai/whisper-1",
  "input": {
    "audio": "https://replicate.delivery/mgxm/e5159b1b-508a-4be4-b892-e1eb47850bdc/OSR_uk_000_0050_8k.wav",
    "response_format": "verbose_json",
    "timestamp_granularities": [
      "segment"
    ],
    "language": "en"
  }
}
```
- Full docs: https://console.core.today/docs/models/openai/whisper-1

### Whisper — `openai/whisper`

OpenAI's Whisper large-v3 speech recognition (144M+ runs) — the standard for transcription. Automatic language detection across ~100 languages, English translation, and plain text / SRT / VTT output formats.

- Provider: OpenAI
- Credits: 14 (per minute of audio (0.2325/s, minimum 10 credits, settled on measured length))
- Speed: medium / Quality: ultra
- Required params: audio (string)
- Optional params: transcription=plain text, translate=false, language=auto, temperature=0, patience, suppress_tokens=-1, initial_prompt, condition_on_previous_text=true, temperature_increment_on_fallback=0.2, compression_ratio_threshold=2.4, logprob_threshold=-1, no_speech_threshold=0.6
- Example input:
```json
{
  "model": "openai/whisper",
  "input": {
    "audio": "https://example.com/meeting.mp3",
    "language": "auto",
    "transcription": "plain text"
  }
}
```
- Full docs: https://console.core.today/docs/models/openai/whisper

### Whisper Diarization — `thomasmol/whisper-diarization`

Whisper transcription with speaker diarization (8M+ runs) — returns who said what, with per-segment speaker labels and timestamps. The go-to for meetings and interviews.

- Provider: Whisper
- Credits: 14 (per run)
- Speed: fast / Quality: high
- Optional params: prompt, file_string, file_url, file, num_speakers, translate=false, language
- Example input:
```json
{
  "model": "thomasmol/whisper-diarization",
  "input": {
    "file_url": "https://example.com/interview.mp3",
    "num_speakers": 2,
    "language": "ko",
    "prompt": "코어닷투데이, API, 크레딧"
  }
}
```
- Full docs: https://console.core.today/docs/models/thomasmol/whisper-diarization