Docs
Quickstart
You need an API key and a prepaid balance. Then you change two lines in your OpenAI client.
Processed on a distributed network of vetted consumer GPUs. Do not send personal or sensitive data. Requests default to the sensitive routing class (our own and vetted cloud hardware only). To use the cheaper distributed network, accept the data policy in the console and send X-Pool-Data-Class: public. Read the data policy first.
1. Get a key
- Sign in to the console with your email. You'll set up an authenticator app on first sign-in.
- Accept the data policy under Settings.
- Top up at least $100 in USDC under Billing (Solana).
- Create a key under API keys. It's shown once; store it in your secret manager.
export VOUCH_API_KEY="sk-pool-…"
2. Change the base URL
The only change from OpenAI is the base URL and the key.
from openai import OpenAI
client = OpenAI(
base_url="https://api.vouchcompute.com/v1", # was https://api.openai.com/v1
api_key=os.environ["VOUCH_API_KEY"],
)
# Transcription (sync, files up to 25 MB)
t = client.audio.transcriptions.create(model="whisper-large-v3-turbo", file=open("space.mp3", "rb"))
print(t.text)
# Embeddings
e = client.embeddings.create(model="bge-m3", input=["gm", "wagmi"])
print(len(e.data[0].embedding))
import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://api.vouchcompute.com/v1", apiKey: process.env.VOUCH_API_KEY });
const t = await client.audio.transcriptions.create({ model: "whisper-large-v3-turbo", file: fs.createReadStream("space.mp3") });
const e = await client.embeddings.create({ model: "bge-m3", input: ["gm", "wagmi"] });
curl https://api.vouchcompute.com/v1/embeddings \
-H "Authorization: Bearer $VOUCH_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "bge-m3", "input": ["gm", "wagmi"]}'
3. Send long files to the batch endpoint
Files longer than a few minutes are cheaper and more reliable as async jobs. On the same GPUs, clips under 30 seconds yield about 4x less audio per dollar than long files.
curl https://api.vouchcompute.com/v1/audio/transcriptions/jobs \
-H "Authorization: Bearer $VOUCH_API_KEY" \
-H "Idempotency-Key: $(uuidgen)" \
-H "X-Pool-Data-Class: public" \
-H "Content-Type: application/json" \
-d '{"model": "whisper-large-v3-turbo", "url": "https://example.org/episode-42.mp3", "response_format": "srt"}'
Then poll GET /v1/jobs/{id} or register a webhook. See Batch API.