Dictation glossary

Cloud transcription

Cloud transcription means speech-to-text that runs on a remote server: your app uploads audio, a model in a data center produces text, and the words come back over the network. It avoids storing a multi-gigabyte model on your disk and can run on machines that cannot host a local engine. The trade: you need connectivity, you accept a round trip, and audio leaves the device for the duration of processing.

Get WhisperJot Try the browser demo
In more detail

What is Cloud transcription?

Cloud transcription is how most consumer voice features shipped for years. The client is a thin recorder; the intelligence lives behind an API. That split makes phones and old laptops feel fast, because they are not loading a 2 GB checkpoint. It also creates a queue: if the network is poor, dictation stutters. Providers differ on retention. Some say they discard audio after inference; some train on it. Those policies are contractual, not visible in the waveform, which is why consent and a clear default matter.

For dictation, cloud mode is a hardware escape hatch more than a different kind of speaking. You still press a hotkey and watch text appear. Accuracy depends on the hosted model, not on the word 'cloud.' Real-time factor can look excellent on a quiet fiber link and poor on hotel Wi-Fi. A product that is cloud-only cannot offer offline dictation. A product that is local-first can still include an optional cloud engine for machines under spec.

For writers who speak

Why it matters for dictation

Cloud transcription is the right question to ask before you dictate anything you would not paste into a web form. It is also the right question if your laptop cannot host a local model. Neither answer is moral failure; mixing them up is. Know the default, know whether a switch is sticky, and know whether audio is stored after the request finishes.

In this product

How WhisperJot handles it

Jot Cloud is WhisperJot's optional cloud engine: opt-in behind a first-use consent prompt, used only while selected, with audio never stored on our servers. It skips the local model download and Neural Engine load, which helps older or low-memory machines. Local engines remain the default. Meetings always stay on-device even if Jot Cloud is selected for dictation.

Related terms

Keep reading in the glossary.

On-device transcription

On-device transcription means speech recognition that runs on the computer or phone in front of you, using a model loaded in local memory.

Read On-device transcription →

Offline dictation

Offline dictation means speech-to-text that keeps working with no internet connection because the recognition model runs on your computer.

Read Offline dictation →

Transcription latency

Transcription latency is the time between speech and usable text — first partial, or the finalized insert after you stop.

Read Latency →

Speech-to-text

Speech-to-text is the process of converting spoken audio into written words using a speech recognition model.

Read Speech-to-text →
Questions

Straight answers.

See also: Pricing, Privacy policy.

What is cloud transcription?

Cloud transcription is speech-to-text performed on a remote server. Your app sends audio, the server runs a model, and text returns over the network. It is how many voice features work on hardware that cannot host a large local model. You need an internet connection, and the audio is off-device while it is processed. Retention after that is a provider policy, not a universal rule.

Is cloud transcription faster than on-device?

Sometimes. A datacenter GPU can decode quickly, and a thin client spends no time loading a local model. Network delay can erase that advantage on a bad link. On a recent laptop, a local engine may return short utterances sooner because there is no round trip. Compare real-time factor on your hardware and connection, not the marketing category.

Does cloud transcription store my audio?

It depends on the vendor. Some process in memory and discard; some keep clips for support or training; some offer a retention control. You cannot see the policy from the microphone light. Read the product's own disclosure for that engine, including whether storage is off, time-limited, or used to improve models. Treat 'never stored' as a specific claim, not a default.

One hotkey, any focused app.

Private local transcription by default, with an optional opt-in cloud engine.

Get WhisperJot — $99/year