Aiko, the free local transcription app for Mac and iOS: speed by device, and its limits

The decision in one paragraph

If your audio cannot leave your device — legal, medical, internal HR, anything under a no-cloud policy — Aiko is the right tool. It is free, runs Whisper entirely on the Mac or iPhone, sends nothing to a server, and the developer collects no telemetry. If your audio can leave your device and you have more than an hour or two per week of it, the math grinds against running Whisper on your own laptop. Whipscribe is hosted, returns the transcript in minutes, accepts URLs, and every recording starts with a free preview, with the summary free too in 31 countries or for recordings of an hour or more in most others: no signup, and a pack from $4 opens the full transcript. Whipscribe pricing checked September 2026.

What Aiko actually is

Aiko is a single-purpose app from Sindre Sorhus, the Mac developer behind a long catalog of polished free utilities. It ships on the Mac App Store and the iOS App Store, free, with no in-app purchases. The full feature surface is short:

• Drop an audio or video file in. Aiko transcribes it locally using a bundled Whisper model.
• Output is a flat text transcript with timestamps. Export as plain text or SRT.
• Recording works on iPhone and iPad — capture in the app, transcribe on the device, no upload.
• 99 languages, because that is what Whisper supports natively.
• No account. No cloud. No telemetry. No subscription.

What Aiko deliberately does not do is also short: no speaker diarization, no URL ingestion, no model picker (the bundled Whisper model is fixed by the app version), no batch queue, no DOCX or JSON export, no API, no MCP, no team workspace. The product does one thing and does it cleanly. That restraint is the appeal.

What Whipscribe is

Whipscribe is a hosted transcription service — a web app, an API, and an MCP server; recordings are transcribed on our own servers and never used for training. The trade is the inverse of Aiko: the audio has to reach a server, and in exchange the transcript comes back in minutes rather than real-time, and the pipeline accepts a URL or a file. Packs from $4, credits never expire. After that, one-time credit packs cost $4 for 300 minutes, $18 for 3,000 minutes or $32 for 6,000 minutes, and credits never expire. Pricing checked September 2026.

The local-Whisper math (scoped to Aiko's constraint)

Aiko bundles a single Whisper model — the user does not pick a tier. Historically that has been Whisper Large with the bundled checkpoint updated by the developer over time. The wait that comes with running Large locally is the same wait every Mac front-end faces, because the bottleneck is the chip and the encoder, not the app wrapper. Numbers below are clean-audio averages on a 1-hour file, drawn from the Whisper paper plus Apple-Silicon community benchmarks (checked May 2026).

Wait times are Apple-Silicon community medians for Whisper Large; iPhone numbers vary heavily with thermal state and concurrent apps. RAM advice assumes you also want the rest of the OS responsive while transcribing.

Side-by-side feature comparison

The two tools share a model family. Almost everything else differs.

The honest cost picture

The line on a receipt says Aiko is $0 and Whipscribe is a one-time credit pack from $4. That is the easy comparison. The harder comparison is what your laptop is doing while Aiko runs.

"GPU-pinned Mac time" is the load-bearing phrase. While Aiko is running, your Mac's GPU is fully occupied with Whisper. The fans spin, battery drains, other heavy work — video calls, Xcode builds, a browser with 40 tabs — competes for the same memory bandwidth. The real cost of "free" isn't the dollar figure. It's the hours your laptop spends being a transcription rig instead of a laptop.

The crossover is around 5 hours of audio per month. Below that, on a recent Apple Silicon Mac, Aiko's compute cost is invisible — the wait fits inside coffee breaks and the device is otherwise idle. Above 5 hours, Mac time starts to dominate any rational accounting, and an $18 credit pack that frees the laptop becomes the better trade.

Worked example — 30 hours of meeting recordings on an M2 MacBook Air

Take a concrete scenario. You're a founder on an M2 MacBook Air with 16 GB of RAM. You record every customer call and every internal sync, ending up with about 30 hours of meeting audio per month. You want to know who said what and you want exports (so the transcripts go into Notion, Linear, or a CRM).

The Aiko path. Drop each meeting file in, wait roughly real-time for Whisper Large to finish — call it 30 hours of compute spread over the month, mostly during evenings or while the Mac is plugged in. The transcripts come out as flat text, no speaker labels. To split "you said" from "they said," you label by hand or copy the text into another tool. You do not get DOCX or JSON, so any downstream automation has to parse plain text. Total dollars: $0. Total Mac hours surrendered to transcription: ~30 per month. Total post-processing time to add speaker labels manually across 30 hours of meetings: probably 4–8 more hours.

The Whipscribe path. Drop the same file in (or paste a URL if it lives in Zoom or Otter cloud). The transcript comes back in 2–5 minutes per hour of audio, and the summary, key quotes and chapters name who said what. Export to DOCX for Notion, JSON for the CRM, SRT if you ever clip the recording for social. Cost: $18 for the 3,000-minute credit pack — 50 hours of audio, and credits never expire. Mac time surrendered: zero — the GPU work happens on a server while you keep working.

When Aiko is genuinely the right call

To be fair to a well-built free app, here are the four situations where Aiko is the right answer and Whipscribe is not.

1. The audio cannot leave the device. Lawyer-client recordings, therapist session notes, internal HR conversations, source interviews under a confidentiality agreement that names the cloud as off-limits, anything covered by a no-third-party-processor policy. Aiko's no-cloud architecture is the feature, and Whipscribe is the wrong tool here.
2. You're on iPhone or iPad and you want a transcription option. Aiko on iOS is one of the few real ways to run Whisper on a phone or tablet. For voice memos, short field interviews, or capture-and-transcribe on the move, the iOS app is genuinely useful. Whipscribe is a website (and an MCP server), so phone-native transcription is not where it competes.
3. You're fully offline. Field journalist in low-connectivity regions, researcher on a long flight, anyone whose primary failure mode is "no internet right now." Aiko works at 30,000 feet. A hosted service does not.
4. Volume is small and you're on Apple Silicon. A handful of voice memos a week on an M2 with 16 GB of RAM. The wait fits inside the time you'd spend on coffee anyway, and you don't need diarization or batch processing. Aiko is fine here.

When Whipscribe is the right call

Outside the four cases above, the math tips toward a hosted pipeline.

1. You have hours of audio per week. Podcasts, journalist interviews, sales calls, founder backlogs, customer-research sessions. The wait stops being free.
2. You need speaker labels. Aiko produces a flat transcript. Anything where "who said what" matters — interviews, meetings, debate-style podcasts — needs diarization. Whipscribe identifies and names the speakers in the summary, key quotes and chapters, so you can see who said what there; every line of the transcript carries a timestamp.
3. The audio lives on the web. YouTube videos, podcast feeds, Zoom or Otter cloud recordings, anything you want to transcribe by URL rather than first-download-then-drag. Whipscribe accepts URLs directly. Aiko does not.
4. You want exports beyond TXT and SRT. Notion likes DOCX. Pipelines like JSON. Subtitles like VTT. Whipscribe ships all of them.
5. You call transcription from Claude, Cursor, or your own code. Whipscribe has an MCP server (used inside Claude Desktop and Cursor) and a REST API. Aiko has neither — it's a Mac/iOS app, intentionally.
6. You're on Intel hardware or a low-RAM device. The local-Whisper wait stops being tolerable. A hosted GPU is genuinely cheaper than your time.

Honest tradeoffs

Two things to be straight about, because the comparison is not all in our favor.

Aiko's privacy story is real, and ours is different. Aiko sends nothing anywhere. Whipscribe sends your audio to a server, processes it, stores the transcript while you have an account, and applies a retention policy. That's a fundamentally different posture. If your threat model says "no audio leaves my device," Whipscribe is the wrong choice and we will tell you so. The reverse is not symmetric — if your threat model says "no audio sits on my unencrypted laptop," neither tool is sufficient and you need disk encryption first.

Aiko is not a competitor — it's a complement to its own niche. The people for whom Aiko is the right tool are not, in most cases, the people for whom Whipscribe is the right tool. Aiko serves the privacy-required, low-volume, file-only segment cleanly. Whipscribe serves the high-volume, URL-friendly, integration-needing segment. The right answer to "Aiko or Whipscribe?" depends entirely on which segment you're in.

The verdict

Aiko is one of the cleanest minimalist apps shipping on the Mac App Store. Free, no-cloud, no nonsense. For the narrow audience it serves — privacy-sensitive, low-volume, willing to wait roughly real-time for Whisper to grind through a file on Apple Silicon — it is the right answer. The dollar cost is zero and the developer collects no data on you.

For everyone else — the operator clearing 30 hours of meetings a month, the journalist with a backlog of interviews, the podcaster with weekly episodes, the team that wants speaker-attributed quotes and summaries and a JSON export — your laptop's time is more expensive than an $18 credit pack. Buy the hours, finish the backlog, and your Mac goes back to being a Mac.

Frequently asked