WalkieTalkey 0.1.0

På dansk

Den første udgave af WalkieTalkey: dansk tale til tekst på din egen computer, gratis og open source.

• Hold fn nede på en Mac (højre Ctrl på Windows), tal, og slip. Eller tryk én gang for at tale uden at holde tasten, og tryk igen for at stoppe.
• Talen bliver til tekst på din egen computer med Edda, en dansk talemodel. Appen henter den selv første gang, mens du gør resten klar.
• Telefonappen: hold knappen, eller tryk én gang, og teksten lander i udklipsholderen på din computer. Krypteret fra telefon til computer.
• Kom i gang på få trin: mikrofon, tilgængelighed (Mac) og prøv det. Ingen konto, ingen sporing.
• Appen er endnu ikke signeret af Apple eller Microsoft. På Mac åbner du appen én gang og trykker derefter Åbn alligevel under Systemindstillinger › Anonymitet og sikkerhed. På Windows trykker du Flere oplysninger › Kør alligevel.

In English

The first WalkieTalkey release, built on Handy: Danish dictation on your own computer, free and open source.

• Hold fn on a Mac (Right Ctrl on Windows), talk and release; or tap once to talk hands-free and tap again to stop. WalkieTalkey only needs Accessibility on the Mac, and offers to stop the Globe key's own action with one button.
• Edda v0.2 as the Danish speech model, converted to GGUF and checked against the original, downloaded by the app from the models-v1 release; Ekko v1 Tiny as the fast model for slower computers; Ekko PnC punctuation on the computer.
• Transcription while you talk, piece by piece at pauses, so a long dictation waits about as long as a short one.
• Danish filler words removed, your own words and replacements (Ordliste), and the Udviklerord pack for English developer words, on by default.
• Optional AI cleanup with your own provider and key (off by default), with consent, guardrails and "Vis hvad der blev sendt".
• The phone app at app.walkietalkey.com: tap or hold the button, and the text lands on your computer's clipboard with a notification. Pairing by QR code, end-to-end encryption through a relay that cannot read the audio, no accounts.
• A larger, calmer desktop app with a floating pill and live waveform while you talk, for key and phone dictations.
• History with retention, a menu bar menu with Seneste, Par en telefon and Sæt på pause, managed settings for companies, updates with stable and beta channels.
• The website at walkietalkey.com with direct downloads and the privacy policy.
• Not signed by Apple or Microsoft yet: the first time, macOS 15 and later needs Open Anyway under System Settings › Privacy & Security, and Windows SmartScreen needs More info › Run anyway.

Models v1

Speech models for WalkieTalkey, converted for transcribe.cpp. The app downloads them from this release and checks every file against SHA256SUMS.

Model	Files	Licence
Edda v0.2 by the Alexandra Institute (default)	edda-v0.2-Q8_0.gguf (886 MB)	Apache-2.0: edda-v0.2-LICENSE, edda-v0.2-NOTICE (base model Whisper, MIT)
Ekko v1 Tiny by RyeAI (fast model)	ekko-v1-tiny-Q8_0.gguf (134 MB)	NVIDIA Open Model License, CC BY 4.0 for RyeAI's changes, CoRal terms: ekko-v1-tiny-*
Ekko PnC v2 by RyeAI (punctuation)	pnc.int8.onnx, tokenizer.json, config.json	CC BY 4.0, CoRal terms: ekko-pnc-*

Model cards: edda-v0.2-README.md, ekko-v1-tiny-README.md, ekko-pnc-README.md (conversion, validation, accuracy, training data).

The use restrictions of the CoRal licence travel with all three models: no impersonating a person (including synthetic speech that resembles a specific person's voice), no detecting or inferring a person's identity or characteristics, and the other uses in Attachment A of that licence. Dictation and transcription are fine. All three were trained partly on FT Speech (Folketinget); whether its terms limit companies' use of models trained on it is an open question (decision 18.7 in docs/MODELS.md).

Converted and published by the WalkieTalkey project (Navigent.io ApS). The weights are the originals, only converted (and for Edda and Ekko v1 Tiny quantised to Q8_0).