A drop-in for OpenAI's /v1/audio/transcriptions endpoint. Same model weights. Same response shape. A fifth of the price, ten times the concurrency.
Measured against OpenAI's published spec. We publish raw numbers because they're competitive on every dimension that matters.
Pass the model name in your request — the rest is the same JSON. No version pinning headaches; old aliases keep working.
Same weights OpenAI ships. Best for interviews, podcasts, anything with real noise or multiple speakers. Use this unless you have a reason not to.
Half the latency, half the price, slightly higher WER. The right choice for clean recordings — meeting bots, voicemail, lecture captures with a good mic.
Acorn-tuned MLX build for clip-length audio. Tuned for ≤30s segments where you need a transcript back immediately — voice memo apps, live chunked streams.
If your code uses the OpenAI SDK today, change the base URL and the key. Everything else — payload shape, error codes, retries — is identical.
from openai import OpenAI client = OpenAI( api_key=os.environ["OPENAI_API_KEY"], ) resp = client.audio.transcriptions.create( model="whisper-1", file=open("interview.mp3", "rb"), response_format="verbose_json", )
from openai import OpenAI client = OpenAI( api_key=os.environ["ACORN_API_KEY"], base_url="https://api.acorncompute.com/v1", ) resp = client.audio.transcriptions.create( model="whisper-large-v3", file=open("interview.mp3", "rb"), response_format="verbose_json", )
$25 of credit lands in your account at sign-up — about 350 hours of whisper-large transcription. No card, no sales call.