REAL-TIME STT
Real-time transcription, low latency.
Transcribe live audio streams with Whisper V3 Turbo for near-real-time results. 8x faster than Whisper Large V3 with near-identical accuracy. Ideal for live captioning, call center analytics, meeting transcription, and voice-controlled interfaces. Per-minute billing, no monthly commitments.
QUICK START
Integrate in minutes.
from openai import OpenAI
client = OpenAI(
base_url="https://api.runcrate.ai/v1",
api_key="rc_live_YOUR_API_KEY",
)
# Use Turbo for lowest latency
transcript = client.audio.transcriptions.create(
model="openai/whisper-large-v3-turbo",
file=open("audio_chunk.wav", "rb"),
response_format="verbose_json",
timestamp_granularities=["word"],
)
for word in transcript.words:
print(f"[{word.start:.2f}s] {word.word}")AVAILABLE MODELS
Models you can use today.
| Model | Provider | Price | Detail |
|---|---|---|---|
| openai/whisper-large-v3-turbo | OpenAI | $0.02/min | 8x faster, ideal for real-time |
| openai/whisper-large-v3 | OpenAI | $0.045/min | Highest accuracy, 100+ languages |
| mistralai/Voxtral-Small | Mistral | $0.03/min | Strong multilingual, long-form |
WHY RUNCRATE
Built for production.
8x Faster with Turbo
Whisper V3 Turbo processes audio 8x faster than the full model with near-identical word error rates. Sub-second processing for short audio chunks.
Word-Level Timestamps
Get precise timing for every word. Essential for live captioning, subtitle generation, and audio-visual synchronization.
Chunked Processing
Split live audio into short chunks and transcribe each one. Stitch results together for a continuous real-time transcript with minimal delay.
100+ Languages
Automatic language detection across 100+ languages. No need to specify the language parameter for most real-time use cases.
FAQ