An abstract spectrogram: hundreds of fine vertical frequency bins burning in magenta and cyan on black, densest at the right and dissolving to darkness at the left, over a solid rail of low-frequency light along the bottom. One column at the centre is perfectly black, carrying no signal at all.

Speech to textlocal by defaultconnected by choice

Run Whisper here, connect to your Shadoword API, or opt into OpenRouter. You choose the path.

Linux releases published · source available.

Read the source

Push to talk

Hold to capture

Default shortcut. Rebindable, or switch to toggle.

LocalDefaultThe default path. Whisper runs on this machine and audio reaches the model without a socket.Shadoword APIAudio is sent to a daemon you host over an authenticated connection you configured.OpenRouterAudio is uploaded after capture stops, only when you explicitly select this cloud path.

Signal path

Nothing istranscribed speculatively.

Capture
Microphone input is recorded and voice activity is detected, so silence never reaches a model.
Segment
Audio accumulates into a segment. Duration, queue length, and idle time are all bounded.
Commit
A commit is what triggers inference. Until one arrives the segment simply sits there.
Deliver
The transcript arrives as text at your cursor, or as a payload to the process that asked for it.

You choose the model and the silicon

Model catalog

Whisper models the desktop client ships in its catalog, with download sizes
Model	Download flag	Size
Large v3 TurboDefaultRecommended balance of speed and accuracy.	`turbo`	1.51 GiB
Large v3Largest and most accurate catalog model.	`large-v3`	2.88 GiB
MediumHigh-accuracy multilingual model.	`medium`	1.43 GiB
SmallBalanced model for lower-memory systems.	`small`	465 MiB
BaseFast model with improved accuracy over Tiny.	`base`	141 MiB
TinySmallest and fastest catalog model.	`tiny`	74 MiB

Accelerator

Build-time accelerator backends and the cargo feature each one needs
Backend	Cargo feature
CPU	`default`Plain cargo build. No feature flag.
Vulkan	`whisper-vulkan`Opt-in cargo feature.
CUDA	`whisper-cuda`Opt-in cargo feature.

The daemon is yours to host

Authenticated, and pointed wherever you point it. Nothing below is illustrative — it is the router and the socket as shipped.

HTTP surface

GET/healthPublic.
GET/v1/status
GET/v1/overview
GET/v1/config
PUT/v1/configRestricted DTO.
POST/v1/transcribe-wavRaw WAV body, 64 MiB cap.
GET/v1/streamWebSocket. Opus packets.
GET/v1/models
POST/v1/models/{id}/select
POST/v1/downloadsExplicit catalog jobs only.
GET/v1/downloads/{id}

Streaming, in order

send{"type":"Start","sample_rate":48000,"channels":1,"protocol_version":2}
recvStartedflow_id, credit
send<binary>raw Opus packets
sendCommitSegmentsegment_index required
recvAcceptedoutstanding, remaining_credit
recvPartialone per committed segment
sendFinish
recvDoneexactly once

Opus in, credit granted at Started and returned at Accepted, one partial per committed segment, one Done at the end.

Build it and run it

Desktop — build and run

cd crates/shadoword-desktop
bun install
bun run tauri dev

Desktop — with Vulkan

nix develop
bun run tauri dev -- --features whisper-vulkan

Daemon — run locally

cargo run -p shadoword-api
# add --features whisper-cuda inside nix develop .#cuda

Daemon — install a model first

cargo run -p shadoword-api -- \
  --download-model turbo

Not yet measured

There is no published latency figure, because no run of the bench corpus has been recorded. When there is one it will name the hardware it ran on.

Corpus: bench_corpus/clip_{10,15,20,30}s.wav
Model: Large v3 Turbo (turbo)