Comparison

Sayboard vs Superwhisper: On-Device Voice Keyboards for iOS Compared

Updated May 2026 ~10 min read

Sayboard and Superwhisper are the closest direct competitors in the on-device AI dictation space on iPhone. Both ship as iOS keyboard extensions. Both run local speech recognition — Superwhisper centers on Whisper, while Sayboard adds Parakeet (NVIDIA) and Moonshine alongside Whisper. Both position themselves around privacy. Where they diverge is on a single architectural choice: whether on-device processing is an absolute, or a default with an opt-in cloud escape hatch.

This page compares them honestly. Sayboard publishes this article — we have tried to keep claims about Superwhisper strictly factual and link to their public sources. Where Superwhisper is the better choice for a given user, we say so.

At a glance

Architecture: pure vs hybrid

This is the structural difference and almost everything else flows from it.

Sayboard is on-device by architecture. The app has no servers. Audio is captured by the keyboard extension, run through a Whisper, Parakeet, or Moonshine model on the device, and the resulting text is inserted into whatever app you are typing in. Optional post-processing — grammar fixes, tone adjustment, formatting — uses smaller LLMs (Gemma, Qwen, Llama, SmolLM2) that you download once and run locally on the iPhone. There is no cloud option to opt into and no way to send your audio elsewhere even if you wanted to.

Superwhisper is on-device by default but explicitly hybrid. Local Whisper models handle transcription offline. On top of that, Superwhisper supports custom "modes" — and Pro modes can include cloud LLM post-processing using user-provided API keys for OpenAI, Anthropic, or other providers. When such a mode is active, transcripts are sent to those third-party services for cleanup. Superwhisper holds SOC 2 Type II and HIPAA certifications.

Both architectures are defensible. Superwhisper's hybrid model is a trade between purity and capability — when you want the polish of a frontier-class LLM, you can have it; when you want offline, you stay offline. Sayboard's pure model removes the choice: privacy is guaranteed by what the app cannot do, not by which mode you toggle.

Open source vs proprietary

Sayboard publishes its full source under GPL-3.0 at github.com/stanlsv/sayboard. Privacy claims are independently verifiable: anyone can read the code and confirm there is no hidden network call. The app can be built from source on a Mac with Xcode at no cost, with the App Store purchase serving as a way to support development.

Superwhisper is proprietary. Privacy guarantees rest on third-party audits — its SOC 2 Type II certification — and on the company's published policies. Users who trust accredited audit bodies more than self-claims will find this sufficient; users who want to read the code themselves will not.

Neither model is strictly better. Open source enables independent verification by anyone with the skills. Certified proprietary architecture provides external review by accountable third parties. The choice depends on which mechanism you trust more.

Pricing models

Sayboard is a one-time App Store purchase with all updates included. No subscription, no recurring fees, no usage cap. Or, since the app is open source, you can build it from source for free.

Superwhisper has three tiers as of 2026:

• Free: 15-minute total recording cap, small local Whisper models, up to 3 custom modes — designed as a permanent trial for evaluation
• Pro: $8.49/month or $84.99/year — larger Whisper models, unlimited custom modes, optional cloud LLM modes via BYOK
• Lifetime: $249.99 one-time — all Pro features forever, all platforms (Mac, Windows, iOS)

The math depends on what you value. For pure offline iOS use, Sayboard's one-time price is dramatically lower than $249.99. For users who want Mac and Windows clients with a synced license and the option of cloud LLMs, Superwhisper's lifetime tier is reasonable — at year three, it breaks even against the annual subscription.

iOS keyboard extension: same form, different parents

This is the part where Sayboard and Superwhisper are most directly comparable. Both ship as system-wide iOS keyboard extensions — meaning either app works inside any text field on iOS, not just inside its own note-taking interface (which is how many other "iOS dictation apps" are constrained).

Both encounter the same iOS limitations: keyboard extensions cannot directly access the microphone in the background, so a brief app-switch is needed to activate the mic. Both have to manage the trade-off between session duration and that switch frequency. The differences here are at the polish level: which one feels more responsive in your typical workflow tends to be a personal-fit question.

Cross-device coverage

Superwhisper runs on macOS, Windows, and iOS with a single account. Settings, custom modes, and license carry across devices. For users who dictate into long-form work on a Mac and into messages on a phone, this is a meaningful day-to-day advantage.

Sayboard is iOS-only. There is no Mac, Windows, web, or Android client maintained by this team. (An unrelated open-source Android voice IME also called Sayboard exists at github.com/ElishaAz/Sayboard by a different developer; it is a separate project.) For users who only dictate on iPhone or iPad, this is a non-issue. For users who live across platforms, Superwhisper covers more surface area.

Custom modes and post-processing

Superwhisper centers its product around custom modes — saved configurations that combine a transcription model with a post-processing prompt and target output style. Pro users get unlimited modes and can use cloud LLMs in those modes for richer cleanup. The result is a highly customizable dictation pipeline: you can have one mode that just transcribes, another that emails-formats output, another that summarizes long-form speech into bullet points.

Sayboard handles the same surface differently. Built-in writing styles (formal, casual, very casual), voice prompts (telling the keyboard what to write), and custom text-processing actions cover most of the same ground. Where Superwhisper has user-defined "modes" backed optionally by frontier cloud LLMs, Sayboard has a more curated set of built-in tools running entirely locally. The customization ceiling is lower; the privacy floor is higher.

Languages and accuracy

Both apps can run Whisper for the underlying speech-to-text layer, so on Whisper-to-Whisper comparisons raw transcription accuracy is comparable. Sayboard also ships Parakeet (NVIDIA) and Moonshine (Useful Sensors) as alternative engines — Parakeet v3 is Sayboard's recommended default. Superwhisper lists 100+ languages with built-in translation to English; Sayboard lists 90+ with automatic detection and also supports speech-to-English translation when using its Whisper models.

For polish on long-form English, Superwhisper's optional cloud LLM modes pull ahead — frontier-class LLMs simply produce smoother output than what a small on-device LLM can match. For offline-only comparison, the gap is much smaller; both rely on local LLMs of similar class for cleanup.

When Sayboard is the better choice

When Superwhisper is the better choice

Verdict

These two products solve the same problem with adjacent philosophies. Sayboard treats on-device processing as an architectural commitment with no exceptions. Superwhisper treats it as a sensible default with a deliberate cloud escape hatch for when you want a more powerful model.

If you read those two sentences and felt clearly drawn to one of them, that's your answer. Most users who pick Sayboard would not feel comfortable with even an opt-in cloud LLM in their voice pipeline. Most users who pick Superwhisper want the option of cloud-class polish for the times they need it, and trust themselves to keep it off when they don't.

A practical heuristic: if you would feel uneasy about an app that could send your transcript to a third-party API even if a setting prevented it, pick Sayboard. If you would not, and you appreciate cross-device coverage and frontier-LLM optionality, pick Superwhisper. Both are credible choices for daily on-device dictation in 2026.

Frequently asked questions

Does Superwhisper work fully offline on iPhone?

Yes for transcription using local Whisper models. No for custom modes that include cloud LLM post-processing — those require an internet connection and send transcripts to external APIs using the user's own keys. Sayboard runs all post-processing locally; there is no cloud option to disable.

Is Sayboard cheaper than Superwhisper?

For pure offline iOS dictation, yes — Sayboard's one-time App Store price is well below Superwhisper's subscription or lifetime tier. For users who want Superwhisper's cross-platform license (Mac, Windows, iOS), the lifetime tier is reasonable per-platform.

Are both apps open source?

Sayboard is GPL-3.0 with full source on GitHub. Superwhisper is proprietary, audited by SOC 2 Type II and HIPAA-compliant rather than open to inspection.

Which one supports more languages?

Superwhisper lists 100+; Sayboard lists 90+ and can additionally translate speech to English via its Whisper models. Both apps can run Whisper for transcription; Sayboard also offers Parakeet (NVIDIA) and Moonshine. Accuracy on common Whisper models is comparable. The gap shows mostly at the long tail of less-resourced dialects.

Does Superwhisper have a Mac and Windows version?

Yes. A single license covers macOS, Windows, and iOS. Sayboard is iOS-only.

Can I use Superwhisper without an account or API key?

Yes for the free tier and for Pro modes that use only local Whisper models. API keys are only needed for cloud LLM modes (BYOK). Sayboard never asks for cloud API keys; all post-processing is local.

Which app produces more polished output?

If you opt into Superwhisper's cloud LLM modes, the output is more polished than any on-device app can match. For offline-only comparison, both apps are close — they run similar-class local LLMs for cleanup.

Also see: Sayboard vs Wispr Flow: AI Voice Keyboards for iOS Compared — for users weighing Sayboard against the leading cloud-based AI dictation product.