Built for professional interpreters, by an interpreter.

Keep the conversation visible.Keep interpreting.

Whisker keeps the client and the person you’re interpreting for visible with live transcription and reference translation—so you can retain context without taking constant notes, while staying in control of the interpretation.

Built to replace the notebook—not the interpreter.

Run locally with Whisker Engine, or connect your own Soniox, Deepgram, or Speechmatics account.

See the workflow in action

Choose a demo

Pick a setup above, then click the recording to start.

Ready

Whisker Interpreter Mode waiting in Standby

The interpreter workflow

One conversation. Two sides. Kept in context.

Generic captions flatten every voice into a single stream. Whisker keeps Client and LEP separate, with the live transcript in Native (original language) and a reference translation in Target (the language you're going to interpret to)—so the conversation stays readable without taking control away from the interpreter.

Client / LEP
Each side keeps its own recent context instead of collapsing every voice into one running transcript.
Native / Target
Native is the original language transcript. Target is a live reference translation—use it to check unfamiliar terms while interpreting in your own voice.
Routing stays yours
Manual routing is the default. Automatic routing is optional when the selected provider supports diarization.

Whisker Interpreter Mode showing separate Client and LEP panels, each with Native transcription and Target reference translation — Local · ManualReal Whisker output

Real Whisker output · Native and Target shown at working size

Reference translation

Translation when you need it. Judgment stays yours.

Target stays beside the confirmed Native transcript as a live second-language reference. Use it to check an unfamiliar term or formulation, then continue interpreting in your own voice.

Confirmed source: Target follows confirmed Native text—not provisional words that are still changing.
Correctable: Edit confirmed Native text when a term is missed and the reference follows the corrected source.
Optional layer: Keep Target visible for support or collapse it per side while transcription continues.

Whisker displays a reference. It does not make the interpretation.

The Whisker stack

Local first. Cloud optional.One workflow, either way.

Transcription and translation are separate choices. Run them on your computer, or use the cloud where you choose.

Choose where audio runs. Keep the translation boundary clear.

Whisker Engine: Turns speech into text on your machine without needing a cloud transcription account.*
Whisker Translator: Produces a live reference translation locally from confirmed words.
Why local: In Local mode, the call stays on your computer and live content is temporary by design.
Privacy by design: No saved transcripts, audio, or clip recordings. Whisker receives no call data; the only server connection is for licensing.

*Real-time local transcription requires a modern GPU. Cloud audio goes only to your selected provider; translation stays local unless cloud translation is selected.

From audio to TargetOne workflow, your choice

INPUTAudio

The live conversation.

TRANSCRIPTION

LOCALWhisker Engine

Whisker's own speech-to-text, running locally on your computer.

CLOUDYour provider

Cloud speech-to-text from Soniox, Deepgram, or Speechmatics.

SOURCENative

The live source text, from provisional to confirmed.

TRANSLATION

LOCALWhisker Translator

Compatible with every transcription source.

CLOUD · WHEN SUPPORTEDCloud Translator

Optional provider-native Target translation.

REFERENCETarget

The language you're interpreting into.

Choose the engine. Control the cost. Compare the difference.

Local or cloud

Your way of working stays consistent. The choice stays yours.

Choose the engine that fits the call: Whisker Engine on your machine, or your own Soniox, Deepgram, or Speechmatics account in the cloud.

Local: Run Whisker Engine, our own whisper.cpp build, on your machine without a cloud transcription account.*
Cloud: Connect Soniox, Deepgram, or Speechmatics with your own API key.
Control: Change the transcription engine without changing the Interpreter workflow.

*A modern GPU is required for real-time local transcription.Validated hardware

Whisker provides the platform—not the minutes.

Cloud usage is billed directly by the selected provider.

Whisker Options · Provider configurationSupported provider example · Soniox

Choose the tool that leaves the decisions to you.Your workflow. Your control.

A subscription-first interpreter bundles the provider, the translation layer, and the meter. Whisker separates them so you can choose what runs locally, what runs in the cloud, and what you actually pay for.

The advantage is not another model. It is control over the whole call.

A capability comparison between subscription-first tools and Whisker
Decision	Subscription-first tools	Whisker
Engine	A bundled cloud path	Your cloud provider or Whisker Engine on GPU
Translation	Dependent on the provider layer	Native translator included; provider translation optional
Connection	Built around an always-available service	Disconnected off-call, closed on hold, and resumed automatically
Workflow	Tied to one product stack	One Client / LEP workflow across engines
Meter	Subscription economics plus provider costs	No Whisker per-minute meter; provider billing stays yours

Same professional standard. More control where it matters. Native translation appears in about 10 ms after the source is confirmed.

Whisker provider costs

Run locally or bring your own cloud account.

Whisker Engine on GPU + Whisker translation$0 / h

Soniox transcription + Whisker translation~$0.12 / h

Soniox transcription + translation~$0.18 / h

Deepgram Nova-3 multilingual + Whisker translation~$0.35 / h

Speechmatics Melia 1 · realtime beta + Whisker translationCost TBA

7-day trial. Use eligible provider credits while they last. After that, the provider bills your account directly—Whisker does not add a per-minute meter. Public provider rates vary by model, language, and usage.

Other apps using Soniox

Published plans converted to effective active-hour pricing.

Subscription app AEffective hourly pricing

150 h · $30 monthly~$0.20 / h

Credit pack · 7 h~$0.71 / h

Credit pack · 60 h~$0.67 / h

Subscription app BEffective hourly pricing

$10 · 22 h~$0.45 / h

$25 · 56 h~$0.45 / h

$100 · 233 h~$0.43 / h

Effective rates are calculated from public plan prices and included hours.

Provider credits

Start with the credits you already have.Keep the provider choice yours.

Connect your own provider account and use any eligible welcome credits from day one. The 7-day Whisker trial gives you time to test the full workflow before paying out of pocket.

ProviderWelcome creditWhisker pathEstimated active hours

Deepgram$200 welcome creditNova-3 multilingual~571 h*

Speechmatics$100 welcome creditMelia 1 · realtime betaTBA

Soniox———

Provider sitesDeepgram Speechmatics Soniox

*Deepgram estimate: $200 ÷ ~$0.35/hour for Nova-3 multilingual. Credit eligibility, expiration, and final rates are set by each provider. Speechmatics Melia 1 realtime beta pricing is currently TBA.

Explore the full workflow before spending a cent out of pocket.

Public release

Whisker 1.0 is coming soon.

Whisker is currently in final development. Availability and pricing will be published here when the first public release is ready.

Current statusFinal development

Keep the conversation visible.Keep interpreting.

One conversation. Two sides. Kept in context.

Client / LEP

Native / Target

Routing stays yours