Available now
Speech transcription
Pilot · speech transcription is live
Async AI jobs without renting GPUs.
Acorn turns API requests into queued work, dispatches them to eligible idle Apple Silicon Macs, and bills per completed job — not per hour of reserved capacity. Transcription is available today; more model routes are planned as the fleet grows.
API surface
OpenAI-compatible + native jobs
Compute
Idle Apple Silicon fleet
Billing
Per completed job
What Acorn is
A queue, a router, and Macs that are already paid for.
Most inference is bursty. Renting GPUs by the hour for that shape leaves capacity idle and bills high. Acorn matches jobs to machines that are already on and free, then settles both customer and host from the same completed-work record.
Submit once. Get a result back.
Post audio to an OpenAI-compatible transcription route or the native async jobs API. Poll for status or receive a signed webhook when the job finishes. No instances, autoscaling groups, or GPU quotas to manage.
Dispatched to eligible Macs.
The coordinator matches each job to an idle Apple Silicon worker with the right model capability, trust tier, and availability. Workers claim only work they can run.
Pay for completed work.
Transcription ships as Fast and Standard SKUs with fixed rates per audio-hour, snapshotted when work is claimed. Volume above 1,000 audio-hours per month is handled by contacting us.
How it works
Submit. Dispatch. Settle.
The same lifecycle for every job: authenticate the request, find an eligible worker, verify the result, and bill from completed work.
01 · Submit
You send the job.
Post audio to a transcription endpoint or the native jobs API. Synchronous routes return when work is ready; async routes return a job id for polling or webhooks.
02 · Dispatch
The router selects a worker.
Acorn matches the job to an idle Mac with the right model, trust tier, and availability. Workers receive only the scoped payload needed for the claim.
03 · Settle
The result is verified and billed.
Completed output returns to your app and is recorded in the ledger. Customers pay per resolved unit; workers are paid from the same completed-work record.
Who Acorn is for
Buyers and hosts share one marketplace.
API customers get capacity without renting hardware. Mac owners get paid for cycles they are not using. Both sides stay optional and reversible.
For API customers
Transcription today. Docs on one page.
- OpenAI-compatible
/v1/audio/transcriptionsplus a native async jobs API. - Fast and Standard SKUs with published per-audio-hour rates.
- Pilot access is operator-reviewed; credentials are issued after approval.
- Signed webhooks, stable error codes, and a single-page reference.
For Mac hosts
Put idle Apple Silicon to work.
- Download the signed macOS app and earn on speech transcription jobs while your Mac is idle.
- You choose when it runs; it pauses on activity, unplugged power, and thermal limits.
- Cash payouts via operator-reviewed PayPal or Venmo settlement — not points.
- Alpha caveat: earnings are limited while the fleet is small.
Trust
Built for scoped work, not full-disk access.
Workers receive only the job payload they need. Legal and security pages spell out permissions, retention, and reporting without marketing soft-pedaling.
Security model
See how the Mac app requests permissions, what the threat model assumes, and how to report vulnerabilities.
Privacy & terms
How data is collected, retained, and deleted during beta, plus the terms that govern API clients and hosts.
An API for jobs, not instances.
Request pilot access for transcription, or put an idle Mac on the network. Both paths are reversible.