What our free tier actually includes — a free dictation app, not a teaser
The free tier ships with Whisper Base — the compact OpenAI Whisper model. On a 2023+ Mac with Apple Silicon or a modern Windows laptop, it runs faster than real time, transcribes English clearly, and handles common vocabulary without a hitch. See how it works for the local-processing architecture.
No time limit per recording. Dictate a one-minute email, a five-minute brainstorm, a fifteen-minute brain dump — the recording runs until you press the hotkey again.
No word cap per month. Dictate as much as you want.
No account required. Download the app, run the installer, set your hotkey, start dictating. We do not ask for an email address until you decide to upgrade to Pro.
No audio uploaded anywhere. The speech model runs locally. Audio is processed in RAM and discarded the moment text is ready. Verifiable in Little Snitch (Mac) or GlassWire (Windows) — no outbound traffic during dictation. See the offline dictation page for the full architectural breakdown.
What Pro adds — and why most people stay on free
Pro is $7.49 / month or $89 / year. It adds the larger Whisper models (Medium, Large v3 Turbo) which materially improve accuracy on technical jargon, accented English, and noisy environments. It adds Parakeet TDT v3 for 25+ European languages with low latency. And it adds the custom dictionary and snippets — the way you teach SnailText words it does not know (your stack, your colleagues, your jurisdiction-specific terms) and short triggers that expand into longer phrases.
Most people who download SnailText stay on the free tier. The compact Whisper Base model is good enough for everyday English dictation. Pro is the upgrade for people whose dictation is technical, multilingual, or done in noisy environments — or who want the dictionary feature to stop fixing the same misheard words manually every day.
There is no "limited free trial" mechanic. The free tier is not a 14-day countdown. It runs as long as you want, no email harvesting, no upgrade prompts blocking core features.
Why "free + local" is rare
Most free voice to text tools are free because they monetize differently — ads, harvesting your audio for model training, freemium upgrade pressure, or selling enterprise plans on top. The cost of cloud STT for the vendor is real, which is why every cloud-based "free" tier has a quota.
Local Whisper has zero per-transcription cost for the vendor. The model runs on your computer, using your electricity and your CPU/GPU cycles. That removes the constraint that forces other "free" tiers to limit you.
The catch is that local Whisper needs a reasonably modern computer. On a 2014 laptop without GPU acceleration, even compact Whisper Base will be slow — closer to 2-3× real time on CPU alone, which is barely usable for live dictation. The hardware floor for usable local dictation in 2026 is roughly: any Apple Silicon Mac (M1+) — see voice to text on Mac, any Windows laptop from 2020 onward with at least an Intel iGPU supporting Vulkan — see voice to text on Windows, or any desktop with a GTX 10-series or newer NVIDIA GPU. For comparisons across the whole category see the best dictation app 2026 roundup.