# Incredibly Fast Whisper - Core.Today AI API

> 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
- **Model ID**: vaibhavs10/incredibly-fast-whisper
- **Category**: Audio & TTS
- **Credits**: 10 per run
- **Speed**: Fast
- **Quality**: Ultra

## Features
- ~150 minutes of audio transcribed in under 100 seconds
- Whisper large-v3 accuracy
- Chunk or word-level timestamps
- Transcribe or translate task modes
- Batched inference with adjustable batch size

## Use Cases
- Long-form podcast and lecture transcription
- Bulk audio archive processing
- Near-real-time transcription pipelines
- Word-level caption alignment
- Call recording analysis

## API Endpoint

Base URL: https://api.core.today/v1
Create Prediction: POST /predictions
Get Status: GET /predictions/{job_id}

## Authentication
Header: X-API-Key: YOUR_API_KEY

## Input Parameters

### Required
- **audio**: string - Audio file to transcribe

### Optional
- **task**: string (default: transcribe) - transcribe (same language) or translate (to English)
  Options: transcribe, translate
- **language**: string (default: None) - Language of the audio ('None' auto-detects; 'korean' for Korean)
  Options: None, afrikaans, albanian, amharic, arabic, armenian, assamese, azerbaijani, bashkir, basque, belarusian, bengali, bosnian, breton, bulgarian, cantonese, catalan, chinese, croatian, czech, danish, dutch, english, estonian, faroese, finnish, french, galician, georgian, german, greek, gujarati, haitian creole, hausa, hawaiian, hebrew, hindi, hungarian, icelandic, indonesian, italian, japanese, javanese, kannada, kazakh, khmer, korean, lao, latin, latvian, lingala, lithuanian, luxembourgish, macedonian, malagasy, malay, malayalam, maltese, maori, marathi, mongolian, myanmar, nepali, norwegian, nynorsk, occitan, pashto, persian, polish, portuguese, punjabi, romanian, russian, sanskrit, serbian, shona, sindhi, sinhala, slovak, slovenian, somali, spanish, sundanese, swahili, swedish, tagalog, tajik, tamil, tatar, telugu, thai, tibetan, turkish, turkmen, ukrainian, urdu, uzbek, vietnamese, welsh, yiddish, yoruba
- **batch_size**: integer (default: 24) - Parallel batches. Reduce if you hit memory errors
- **timestamp**: string (default: chunk) - Chunk-level or word-level timestamps
  Options: chunk, word
- **diarise_audio**: boolean (default: false) - Use Pyannote.audio to diarise the audio clips. You will need to provide hf_token below too.
- **hf_token**: string - Provide a hf.co/settings/token for Pyannote.audio to diarise the audio clips. You need to agree to the terms in 'https://huggingface.co/pyannote/speaker-diarization-3.1' and 'https://huggingface.co/pyannote/segmentation-3.0' first.

## Examples

### Fast Podcast Transcription
Transcribe a long recording with word timestamps
```json
{
  "model": "vaibhavs10/incredibly-fast-whisper",
  "input": {
    "audio": "https://example.com/podcast-episode.mp3",
    "task": "transcribe",
    "timestamp": "word"
  }
}
```

## Response Format
```json
{
  "job_id": "abc123",
  "status": "pending | processing | completed | failed",
  "result": "URL or data (when completed)"
}
```

## Usage Flow
1. POST /predictions with model and input -> receive job_id
2. GET /predictions/{job_id} -> poll until status is completed or failed
3. Result contains output URL(s)

## Tips
- The fastest way to transcribe hour-plus audio in the catalog
- timestamp: 'word' enables karaoke-style caption alignment
- Same 13-credit price as standard Whisper — pick this for long files
- Reduce batch_size to 12-16 if long files fail with memory errors

## Documentation
https://replicate.com/vaibhavs10/incredibly-fast-whisper