For developers

Speech to text API with Whisper large-v3 turbo

In testing, opens soon

Short answer: Smarty Pants has a speech to text API that runs Whisper large-v3 turbo on Apple Silicon Macs through the OpenAI-format endpoint POST https://smartypants.network/v1/audio/transcriptions. Each file can be up to 30 minutes or 25 MB. The listed price on 8 Nov 2026 was $0.18 an hour of audio. The model is in testing and opens soon, so it is not for production traffic yet.

Smarty Pants is testing a speech to text API that runs Whisper large-v3 turbo on Apple Silicon Macs. It uses the same OpenAI-compatible API as the chat models, so existing transcription code needs only a new base URL and key. It is in testing and opens soon.

What is the Smarty Pants Whisper API?

It is OpenAI’s open Whisper large-v3 turbo model, served on Macs behind POST /v1/audio/transcriptions. Right now it is in testing: the live docs list it as “early testing, not for production traffic” (llms.txt).

Spec Value
API name smartypants/whisper-large-v3-turbo
Endpoint POST https://smartypants.network/v1/audio/transcriptions (multipart upload)
Limit per file up to 30 minutes or 25 MB
Listed price (8 Nov 2026) $0.18 an hour of audio
Build Core ML
License MIT
Served on qualified 16 GB+ Apple Silicon Macs
Status In testing, opens soon

The Hugging Face model card describes large-v3 turbo as a pruned Whisper large-v3 with 4 decoding layers instead of 32: “way faster, at the expense of a minor quality degradation.” It lists support for many languages and predicts the source language automatically.

How do you send audio?

Upload the file as multipart form data with your Smarty Pants key, in the OpenAI format. Anyone can sign up with an email at smartypants.network/api and create a key; new accounts get $10 of free credit, usable for 90 days.

curl

curl https://smartypants.network/v1/audio/transcriptions \
  -H "Authorization: Bearer $SMARTYPANTS_API_KEY" \
  -F model="smartypants/whisper-large-v3-turbo" \
  -F file="@meeting.m4a"

Python

import os
from openai import OpenAI

client = OpenAI(base_url="https://smartypants.network/v1", api_key=os.environ["SMARTYPANTS_API_KEY"])

with open("meeting.m4a", "rb") as audio:
    transcript = client.audio.transcriptions.create(
        model="smartypants/whisper-large-v3-turbo",
        file=audio,
    )
print(transcript.text)

While the model is in testing, is_ready on /v1/models may be false and requests may return 429 or 503. Wait the Retry-After seconds or send the file to your current provider. You are not charged for a request Smarty Pants can’t serve.

What does an hour of audio cost?

The listed price on 8 Nov 2026 was $0.18 an hour of audio (smartypants.network/api). Prices can change; /v1/models has the current one, and /inference/pricing/ has a calculator.

That is not the lowest listed price for this model. On 9 Nov 2026 Groq’s supported models page listed Whisper Large V3 Turbo at $0.04 an hour, with files up to 100 MB. If price per hour is what matters most, compare both before choosing. See Smarty Pants vs. centralized providers.

What is it good for?

Turning recordings into text you can search, summarize or label.

  • Meeting and call notes, then a summary with a chat model.
  • Draft captions for podcasts and videos.
  • Voice memos and interviews you want to search.

Keep files under 30 minutes or 25 MB each. For a long recording, split it into parts and join the text afterward.

What should you not upload?

Recordings with sensitive content. Requests run on Macs other people own, and a serving Mac must decrypt a request in memory to answer it. The live site says: “General network: please don’t send sensitive data.” Calls with customers, patients or private conversations belong on a provider you have approved for that data.

Frequently asked questions

Can I use the Smarty Pants Whisper API today?

Not for production. Whisper large-v3 turbo is in testing and opens soon; the live docs mark it "early testing, not for production traffic." Check is_ready on /v1/models.

How long can an audio file be?

Up to 30 minutes or 25 MB per file. Split longer recordings into parts before you upload them.

What does it cost?

The listed price on 8 Nov 2026 was $0.18 an hour of audio. Check /v1/models or smartypants.network/api for the current price.

Does it work with the OpenAI SDK?

Yes. The endpoint uses the OpenAI format with a multipart audio upload, so client.audio.transcriptions.create works once you set the Smarty Pants base URL and key.

Which version of Whisper is it?

Whisper large-v3 turbo, served as a Core ML build. The model card on Hugging Face describes turbo as a pruned large-v3 with 4 decoding layers instead of 32, faster with a minor quality loss.

Sources

  1. Smarty Pants: llms.txt
  2. Smarty Pants: Agent kit and docs
  3. Smarty Pants: GET /v1/models
  4. Smarty Pants: API guide (listed audio price)
  5. Hugging Face: openai/whisper-large-v3-turbo model card
  6. GroqCloud: Supported models (checked 9 Nov 2026)

Try a pinned open model on real Macs.

See every model, its listed price and the code to call it. Smarty Pants works with the OpenAI SDK you already use: change the base URL and the key.

Requests run on Macs owned by other people. Owners agree not to read requests and the app doesn't show them, but a serving Mac decrypts each request in memory to answer it, so that is a rule and a design, not a technical guarantee. Don't send sensitive data on the general network.