Comparison
Sayboard vs Wispr Flow: AI Voice Keyboards for iOS Compared
Both Sayboard and Wispr Flow turn speech into text on iPhone, and both use modern AI to do it. From there they diverge sharply: where the audio is processed, how you pay, what happens to your voice after the words appear on screen. This page lays out the differences honestly, so you can pick the one that fits your work and your tolerance for cloud processing.
This comparison is published by the Sayboard team. We have tried to keep claims about Wispr Flow strictly factual and link to their public sources. Where we think Wispr Flow is the better choice for a given user, we say so.
Sayboard runs entirely on the iPhone — no servers, no internet required after the initial model download, one-time purchase, open-source under GPL-3.0.
Wispr Flow runs in the cloud — your audio is sent to remote servers for transcription and LLM-based cleanup, charged as a subscription, with optional Privacy Mode that prevents retention.
If you want privacy by architecture and a one-time price, pick Sayboard. If you want the most polished cloud-LLM output, multi-platform sync, and named compliance certifications, pick Wispr Flow.
At a glance
| Sayboard | Wispr Flow | |
|---|---|---|
| Where audio is processed | On your iPhone or iPad | Cloud servers |
| Internet required | Only for initial model download | Yes, for every dictation |
| Pricing | One-time App Store purchase | Subscription: $15/mo or $144/yr (Pro). Free tier: 2,000 words/week. |
| Open source | Yes (GPL-3.0) | No |
| Languages | 90+ with automatic detection; speech-to-English translation via Whisper models | 100+ with cross-language switching |
| Speech models | Whisper, Parakeet, Moonshine (on-device) | Cloud LLMs (transcription + post-processing) |
| iOS launch | March 2026 | May 2025 |
| Compliance certifications | None — no servers to certify | SOC 2 Type II, ISO 27001, HIPAA BAA available |
| App Store rating (May 2026) | Not enough ratings to display | 4.8 stars (8,000+ reviews) |
| Subscriptions / accounts | None | Account required |
Privacy: where your voice actually goes
This is the structural difference, and it determines almost everything else.
Sayboard performs speech recognition on the device itself. Audio is captured by the keyboard extension, processed by a Whisper, Parakeet, or Moonshine model running locally, and the resulting text is inserted into whatever app you are typing in. The audio never leaves the phone. Sayboard has no servers — there is literally no infrastructure to send anything to. The only network call the app makes is the initial download of a speech model from a CDN, which is a standard file fetch with no account, no identifier, and no usage logging.
Wispr Flow performs transcription in the cloud. According to their public documentation, audio is uploaded to remote servers (which use models from companies like OpenAI and Meta), transcribed there, and the cleaned-up text is returned to the app. Wispr offers a "Privacy Mode" that, when enabled, prevents storage of audio and transcripts after processing and excludes data from model training. Privacy Mode is available on the free tier. Wispr Flow also publishes SOC 2 Type II and ISO 27001 reports and offers a HIPAA Business Associate Agreement.
The honest framing: Wispr Flow has a serious, certified privacy posture for a cloud product. Sayboard's posture is different in kind — when audio cannot leave the device, the question of "what happens on the server" is moot. Privacy Mode is a policy guarantee; on-device processing is an architectural one.
Pricing: one-time vs subscription
Sayboard is a one-time purchase on the App Store. Pay once, keep the app forever. All future updates — new speech models, new features, new language support — are included. There is no subscription, no annual renewal, no usage cap. Because Sayboard is also open-source under GPL-3.0, you can alternatively build it from source with a Mac and Xcode at no cost; the App Store purchase exists to support development.
Wispr Flow is subscription-based. As of 2026, the published tiers are:
- Basic (free): 2,000 words per week cap, full Privacy Mode
- Pro: $15/month month-to-month, or $12/month billed annually ($144/year)
- Teams: $12/user/month month-to-month, or $10/user/month billed annually, three-seat minimum
- Enterprise: custom pricing
New accounts get a 14-day Pro trial without a credit card. The free tier is genuinely usable for low-volume dictation — if you only dictate a few short messages a week, you can stay on it indefinitely.
The math over time: a Wispr Flow Pro annual plan at $144/year exceeds Sayboard's one-time price within the first month for most users. Over five years of use, the gap widens substantially. For high-volume professionals, the subscription buys cloud LLM polish; for everyone else, the math favors a one-time purchase.
Accuracy and output quality
Both apps transcribe well. The differentiator is what happens after transcription — the post-processing layer that removes filler words, fixes grammar, and formats text for context.
Wispr Flow runs cloud-scale large language models in this layer. The result is consistently polished output: "um" and "uh" are removed, run-on sentences are split, code comments are formatted as code, emails are formatted as emails. For users who dictate long-form text into Slack, email, or documents and ship it without editing, this is the biggest day-to-day quality advantage of the cloud architecture.
Sayboard runs smaller LLMs locally — Gemma, Qwen, Llama, SmolLM2 are available as optional downloads. Post-processing exists but is more conservative; the model has fewer parameters and works within the iPhone's RAM constraints. Raw transcription accuracy is competitive (Whisper-class models are state-of-the-art for on-device English transcription), but the cleanup layer does less. If you tend to read back and edit before sending, Sayboard's output is fine. If you want to publish dictation un-edited, Wispr's polish is noticeable.
Offline support
Sayboard is fully offline after the initial model download. Airplane mode, subway tunnel, plane without Wi-Fi, paid hotel internet, sketchy mobile connection — none of it matters. The keyboard transcribes locally regardless.
Wispr Flow requires an active internet connection for every dictation. There is no offline mode. If your network drops mid-sentence, transcription pauses or fails. For users who dictate frequently while traveling or in network-constrained environments, this is a meaningful day-to-day difference.
Open source and self-hosting
Sayboard publishes its full source code under GPL-3.0 at github.com/stanlsv/sayboard. Anyone can audit the code, build the app from source, fork it, or contribute. For privacy-conscious users this is verification: the privacy claim ("audio never leaves the device") is checkable, not just promised.
Wispr Flow is proprietary. Privacy claims rely on their published policies, certifications, and Privacy Mode behavior. There is no source code to audit; the architecture is what their security team and certifications say it is.
Both models are legitimate. Open source enables independent verification; certified proprietary architecture provides external audit by accredited bodies. Which one you weigh more depends on whether you trust source-code transparency or third-party auditors more.
Platforms and ecosystem
Sayboard is iOS-only — iPhone and iPad with iOS 17 or later. There is no Mac, Windows, web, or Android version maintained by this team. (An unrelated Android keyboard called Sayboard exists at github.com/ElishaAz/Sayboard; it is a separate project by a different developer.)
Wispr Flow ships on macOS, Windows, and iOS, with a shared account that syncs settings and customizations across devices. For users who dictate into long-form work on a desktop and into messages on a phone, the cross-device experience is smoother.
When Sayboard is the better choice
You care about privacy by architecture
You want a guarantee that audio cannot reach a third party because it has nowhere to go, not because a policy says so.
You don't want a subscription
One-time purchase, no recurring fees, no cap on usage, no risk of price hikes or sunset.
You dictate offline often
Travel, commutes, planes, low-signal areas — Sayboard works the same everywhere after the model is downloaded.
You value open-source verification
You want to read or audit the code yourself, or you want to be able to build the app from source.
When Wispr Flow is the better choice
You ship dictation un-edited
Cloud LLM cleanup is meaningfully more polished than what fits on-device. If you dictate emails or documents and send them as-is, this matters.
You work cross-device
Mac and Windows clients with synced settings. Sayboard is iOS-only.
You need named compliance
SOC 2 Type II, ISO 27001, HIPAA BAA. If your compliance team requires named certifications, this is the cleaner path.
You want to try free first
The 2,000 words/week free tier and 14-day Pro trial let you evaluate without committing.
Verdict
These products solve the same problem with opposite architectures. Sayboard solves it on-device; Wispr Flow solves it in the cloud. Each approach has real, non-rhetorical advantages. The cloud approach delivers more polished output and broader platform support. The on-device approach delivers structural privacy, no subscription, and offline reliability.
Most "AI voice keyboard" comparisons frame this as a winner-takes-all matchup. It is not. They are different tools optimized for different priorities, and there is no single right answer.
That said, here is a practical way to decide: if you would feel uneasy about your voice being uploaded to a third-party server even with a Privacy Mode flag enabled, pick Sayboard. If you would not, and you value polished output and cross-device support, pick Wispr Flow. Either is a credible choice for daily dictation in 2026.
Frequently asked questions
Does Wispr Flow work offline on iPhone?
No. Wispr Flow requires an internet connection on every dictation. All audio is sent to cloud servers for transcription and post-processing — there is no on-device mode.
Is Sayboard cheaper than Wispr Flow?
Over almost any usage horizon longer than a couple of months, yes. Sayboard is a one-time App Store purchase with all updates included. Wispr Flow Pro is $15/month or $144/year. Wispr's free tier is genuinely usable for low-volume dictation, so for very light use it can be the cheapest option.
Which one is more accurate?
Raw transcription accuracy is competitive on both. Wispr Flow's cloud-scale post-processing produces more polished output for un-edited long-form dictation. Sayboard's on-device cleanup is more conservative but sufficient for most workflows where you read back before sending.
Can I use Sayboard without paying?
Yes. The full source is published under GPL-3.0; you can build the app yourself with a Mac and Xcode at no cost. The App Store purchase is a way to support development.
Does Sayboard send any data to its developer?
No. Sayboard has no servers, no analytics, no telemetry. The only network call is the initial download of a speech model. Wispr Flow sends every audio clip to its cloud, with the option to enable Privacy Mode so the data is not retained or used for training.
Which app supports more languages?
Wispr Flow lists 100+ languages; Sayboard lists 90+ and can additionally translate speech from any of them to English via its Whisper models. Both cover all major European and Asian languages and most regional dialects. The practical gap is small.
Which is better for HIPAA or regulated work?
Wispr Flow has named certifications (SOC 2 Type II, ISO 27001) and offers a HIPAA BAA. Sayboard has no certifications because it has no cloud infrastructure to certify — your audio never leaves the device. Both can be defensible answers to a compliance review, depending on what the policy requires.
Also see: Sayboard vs Superwhisper: On-Device Voice Keyboards for iOS Compared — for users weighing Sayboard against the closest pure-on-device competitor.
Try Sayboard
iPhone and iPad. iOS 17 or later. One-time purchase. Open-source.