EchoWrite

EchoWrite is a Windows (and macOS) dictation tool that sits system-wide, turning speech into text inside whatever application has focus — a browser field, a document, a terminal, an email. It is bring-your-own-key (BYOK) for cloud transcription models, plus a documented local-processing option that runs an offline model on the machine instead of sending audio to a vendor's servers. The vendor's own homepage copy sums up the pitch in four words: "Free · No data collection."

That line is also exactly the claim buyers need to be most careful about. A dictation tool by definition captures everything you say near an open microphone — drafts, passwords read aloud, client names, anything. Whether a "local" mode genuinely keeps that audio on the device, or merely processes it locally before something still phones home, is not something a reader can tell from a landing page. It can only be shown by watching the network traffic while the software runs.

We could not do that here. TTML's current testing environment for this product is a headless Linux container with no Windows or macOS host, no display server, no audio device and no way to install or run the application. We have verified what EchoWrite's site says. We have not verified what the software does. That distinction runs through every section below.

What it actually is

EchoWrite is a native desktop application for Windows 10/11 and macOS. The documented feature set covers system-wide dictation into any application, filler-word removal, real-time translation, voice shortcuts, and a custom vocabulary/lexicon for names and technical terms. It offers two processing paths: a local mode (an offline model, described by the vendor as Whisper-based or a local LLM) and a cloud mode where the user supplies their own API key to a third-party model provider — hence "BYOK". The vendor markets a lifetime/perpetual entitlement rather than a recurring subscription, at a low headline price.

What we verified

• The vendor site is live and reachable. A headless render returned HTTP 200 and the published claims quoted above, reproduced on both a curl fetch and a browser render.
• Published pricing exists, including a free entry point, alongside paid tiers on the vendor's own site.
• A public developer footprint exists. The person behind EchoWrite has posted about building it in a community channel (Reddit), consistent with an early-stage, independently-built product. We record this as a vendor-adjacent claim, not independent evidence — a first-party post in a public forum is still the vendor talking about themselves.

Nothing above is a hands-on functional test. We have not installed the application, dictated into it, or measured anything it does. Our testing environment cannot run Windows or macOS software, has no display and no microphone, so none of that was attempted — it was not skipped for convenience.

What we could not verify

• Local-processing / no-data-collection claim — not confirmed. Settled only by a network capture during an active local-mode dictation session on real Windows/macOS hardware.
• Transcription accuracy — not confirmed. Requires dictating a fixed test passage in both local and cloud modes and measuring word error rate against the source text; the two modes must be scored and reported separately, never merged.
• Latency — not confirmed. Requires timing the gap between speech ending and text appearing, across a run of utterances in each mode.
• Application compatibility breadth — not confirmed beyond the vendor's own description. Requires dictating into a browser field, a native editor and an Electron-style app and recording per-application results, since compatibility in one text field cannot be generalised to all of them.
• What model(s) ship locally, and what hardware they need — not confirmed. Requires inspecting on-disk model artefacts after installation.
• Where audio and transcripts are stored, and whether that storage is encrypted — not confirmed.
• The vendor's operating entity and support commitment behind a lifetime entitlement — not confirmed. A lifetime price is only as durable as the business selling it, and we found no independent corporate detail beyond the site and one community post.
• Independent, third-party testing of EchoWrite — none located anywhere. Every claim currently traces back to the vendor.

Pricing and entitlement

The vendor's own site publishes a free tier plus paid price points; we have not captured the full tier-to-feature mapping in a structured way, so treat any specific figure quoted elsewhere as unverified until you check the current page yourself. Separately, EchoWrite has appeared as a marketplace listing at $29.00 (a snapshot from a data feed dated 9 July 2026) — that is a marketplace listing price on a given date, not a TTML price guarantee, not a discount claim, and not something we have reconciled against a verified vendor reference price. If EchoWrite is a lifetime/perpetual entitlement, what happens to that entitlement if the vendor stops trading, changes model support, or discontinues local models is not addressed anywhere we could find, and should be treated as an open question rather than assumed to be covered.

BYOK cloud mode carries a real running cost beyond the purchase price: the user pays their own API provider directly for cloud transcription usage. Factor that into any comparison with a flat-fee competitor — EchoWrite's headline price is not the full cost of using its cloud mode.

Where it holds up

The product exists, is reachable, and is transparent about its own architecture in a way some competitors are not: it tells you plainly that there are two modes, that one needs your own API key, and that the other is meant to run locally. A BYOK model is also a genuinely different commercial shape from a fixed monthly subscription, and for a user who already holds API credits or a preferred model provider, that can be a real advantage rather than a workaround.

Where the ceiling is

Everything that would make the privacy pitch trustworthy — the exact thing that differentiates EchoWrite from a plain cloud dictation tool — is currently unconfirmed. So is the accuracy that makes any dictation tool worth using day to day. This is a young, independently-built product with a thin public track record and no located third-party testing. None of that means the product is bad; it means the burden of verification currently sits entirely with the buyer, not with any published evidence.

How it compares

The natural comparison point for system-wide, low-latency dictation is Wispr Flow, and the natural free baseline is Windows' own built-in voice typing — neither currently has a TTML review to link to, so we name them in plain text rather than fabricate a link. The fairer architectural comparison is the wider field of Whisper-based open-source dictation tools, since they share EchoWrite's local-model approach and make the privacy question directly comparable rather than a matter of trusting marketing copy. If your actual goal is a broader AI productivity workflow rather than dictation specifically, it is also worth checking the AI productivity category for tools that have been reviewed with more evidence behind them, such as Notion, where account-based and desktop-app functionality has been assessed against what TTML could actually confirm.