Does on-device dictation work offline?

Yes for the transcription path. After the app is installed and any extra models you chose are downloaded, hold the shortcut and speak without a network. Account status, billing, updates, and optional support still need the internet when you use those features.

Does VeloxWaves do on-device dictation on Windows?

Yes. Local speech recognition is the core path on Windows, macOS, and Linux. It is not limited to Apple Silicon, and Windows is not a cloud fallback.

How is this different from Apple Dictation?

Apple Dictation can process some languages on-device on recent Apple silicon, and it can also send audio to Apple depending on language, length, and settings. VeloxWaves is a dedicated desktop dictation app: push-to-talk into any field, local history, and a private knowledge graph. It also runs on Windows and Linux. See the Apple Dictation alternative page for the full split.

How is on-device different from zero-retention cloud?

Zero-retention cloud still sends audio to a remote model, then promises to delete it. On-device means the model runs on your hardware, so the audio never takes that trip. Both are privacy strategies. They are not the same architecture.

Is this the same as voice typing?

Voice typing, push-to-talk dictation, and local speech-to-text are the same job when the text lands in the app you already have focused. VeloxWaves does that system-wide, then keeps the transcript as searchable local memory. The dedicated voice typing page covers the cursor loop.

On-device dictation

On-device dictation for Windows and Mac

An on-device dictation app turns speech into text on your computer. VeloxWaves does that in any desktop app, then keeps the transcript in a private second brain. Raw voice audio is not uploaded for transcription.

Download free

No credit card. Windows, macOS, and Linux.

Definitions

What on-device dictation actually means

“On-device” is used loosely. Some products run a wake word locally and send the recording to a server. Some transcribe locally and still upload text. Some offer a local model and keep cloud as the default. When VeloxWaves says on-device, the transcription path itself runs on your hardware.

Flavor	What actually happens
Hybrid	A small model may run locally for a wake word or preview. The actual transcription is sent to a server.
On-device, then uploaded	Speech recognition happens locally, but transcripts, analytics, or audio snippets still leave the machine.
Mode-dependent	Local models exist, but the default path is cloud unless you switch on a local-only setting.
VeloxWaves	Capture, transcription, post-processing, and memory all run on your computer. There is no cloud transcription mode to turn off.

That is why VeloxWaves still works when Wi-Fi is off. Compare the architecture, not the marketing phrase. A longer product-by-product view is in our 2026 on-device dictation roundup.

Pipeline

How VeloxWaves transcribes on your computer

From shortcut to inserted text, the speech path stays on the machine. The same loop is the product's answer to voice typing: speak, and the words land where the cursor already is.

Step 1
You hold the shortcut
VeloxWaves opens the microphone only while you hold push-to-talk. Nothing was listening before that moment. Default shortcuts are Ctrl+Win on Windows and Ctrl+Cmd on Mac.
Step 2
Audio stays in memory
Samples are buffered in RAM, including a short pre-roll so the first syllable is not cut off. Raw voice audio is not uploaded for transcription.
Step 3
A local speech model runs
Moonshine is the default English engine. Whisper is available for non-English languages after an explicit download. Hardware tiers choose a model that fits the machine you have.
Step 4
Text is inserted at the cursor
The transcript is typed into the app you were already using. Dictation is a keyboard feature, not a separate chat window you copy from.
Step 5
The words become private memory
Each dictation can feed a local knowledge graph. Semantic search, translation, and on-device answers use that store. None of it is a hosted second brain.

Network

What still uses the network

Local dictation does not mean the app never talks to a server. It means the content path — audio, transcript, embeddings, and local AI prompts — is not a hosted transcription product. The Trust & Privacy guide maps the same split in more detail.

Account and billing

Sign-in, trial status, and subscription checks contact the account service. They do not include dictation audio.

Consented model downloads

Larger speech, translation, or local AI models download only when you ask. Inference after that stays on-device.

App updates

The updater checks for new desktop builds. Release files are product binaries, not your transcripts.

Optional support reports

Logs and screenshots leave the machine only if you submit an issue. Review them before sending.

Platforms

Windows, Mac, and Apple-only tools

Several well-known on-device dictation apps are excellent on Apple silicon and stop there. Others offer a local mode on Mac and a cloud path on Windows. VeloxWaves keeps the same local transcription promise on Windows, macOS, and Linux.

That matters if your work machine is a PC, if a team mixes platforms, or if “on-device” has to keep meaning the same thing after you leave a MacBook. Hardware-aware model selection is how the app stays usable on everyday machines instead of requiring a Neural Engine.

Apple-first products can still be the right choice for iPhone keyboards, App Store workflows, or one-time Mac licenses. Fair comparisons: Dictly, Voibe, Spokenly, and On Device AI.

FAQ

Frequently asked questions

Does on-device dictation work offline?: Yes for the transcription path. After the app is installed and any extra models you chose are downloaded, hold the shortcut and speak without a network. Account status, billing, updates, and optional support still need the internet when you use those features.
Does VeloxWaves do on-device dictation on Windows?: Yes. Local speech recognition is the core path on Windows, macOS, and Linux. It is not limited to Apple Silicon, and Windows is not a cloud fallback.
How is this different from Apple Dictation?: Apple Dictation can process some languages on-device on recent Apple silicon, and it can also send audio to Apple depending on language, length, and settings. VeloxWaves is a dedicated desktop dictation app: push-to-talk into any field, local history, and a private knowledge graph. It also runs on Windows and Linux. See the Apple Dictation alternative page for the full split.
How is on-device different from zero-retention cloud?: Zero-retention cloud still sends audio to a remote model, then promises to delete it. On-device means the model runs on your hardware, so the audio never takes that trip. Both are privacy strategies. They are not the same architecture.
Is this the same as voice typing?: Voice typing, push-to-talk dictation, and local speech-to-text are the same job when the text lands in the app you already have focused. VeloxWaves does that system-wide, then keeps the transcript as searchable local memory. The dedicated voice typing page covers the cursor loop.

On-device dictation for Windows and Mac

What on-device dictation actually means

How VeloxWaves transcribes on your computer

You hold the shortcut

Audio stays in memory

A local speech model runs

Text is inserted at the cursor

The words become private memory