Dictus Pro — Premium features roadmap (Open Core model)

[Pivii]

Model

Open Core — all code stays public (MIT license). Pro features are gated by isPro boolean, unlocked via payment (Bitcoin or Play Store subscription).

Mirrors iOS roadmap: getdictus/dictus-ios#54

Free forever (no paywall)

• Keyboard dictation (Whisper + Parakeet)
• Text suggestions FR+EN with personal dictionary
• Model manager (download, delete, switch provider)
• Onboarding
• AZERTY / QWERTY layouts
• Emoji picker
• Settings (theme, haptics, sound, language)

Pro features

Tier 1 — Core Pro (priority)

Feature	Description
Smart Mode Pro	Intelligent reformulation + templates (email, notes, SMS, summary, bullet points) via local LLM
Custom vocabulary	Personal dictionary of technical/rare terms injected as Whisper initial prompt
Transcription history	Local journal of past dictations with full-text search
Audio file transcription	Import .mp3/.m4a/.wav via share intent, local Whisper transcription
Voice message transcription	Transcribe voice notes from messaging apps

Tier 2 — Advanced Pro

Feature	Description
Contextual reformulation	"More formal", "shorter", "like an email"
Auto-summary	Bullet points from long dictation
Action extraction	TODOs from meeting transcription
Local translation	Dictate FR, get EN (or reverse)

Tier 3 — Power User

Feature	Description
Continuous long dictation	>5 min with chunked transcription
Dictation profiles	"Doctor mode", "Dev mode", "Lawyer mode"
Multi-format export	.txt, .md, .docx

Pricing strategy

Channel	Model	Price
Play Store	Monthly subscription	~2-3€/mo (TBD)
Play Store	Yearly subscription	~20-25€/yr (TBD)
GitHub / F-Droid	One-time Bitcoin payment	~15-20€ equivalent (TBD)

Beta period

During beta, all Pro features are unlocked for free with clear messaging that they will become paid. This lets us validate demand and gather feedback before monetizing.

Technical implementation

• #1 — Gradle product flavors (libre + playstore)
• Gradle product flavors: libre + playstore #2 — ProUnlocker interface + isPro gating
• ProUnlocker interface + isPro feature gating #3 — BTCPay Server integration (Bitcoin, libre flavor)
• BTCPay Server integration — Bitcoin one-time payment (libre flavor) #4 — Google Play Billing (subscription, playstore flavor)

Principles

• 100% offline — no cloud, no data leaves device
• Privacy-first as core differentiator
• No bait-and-switch: free features never become paid
• Open source: anyone can compile and bypass the paywall — that's fine
• Bitcoin community is a primary target audience

[m13v]

solid roadmap. the custom vocabulary feature (Tier 1) is underrated and will be a real differentiator. we built something similar for our voice-first macOS agent and the accuracy improvement for domain-specific terms is dramatic. Whisper's initial_prompt parameter lets you prime the model with expected vocabulary, and for medical/legal/tech terms it goes from unusable to reliable.

for the audio file transcription feature, one implementation detail that matters: process the audio in chunks rather than feeding the whole file at once. Whisper models have a 30-second context window, and for longer recordings you get better results with overlapping 30-second segments that you stitch together. without the overlap, you lose words at segment boundaries.

for continuous long dictation (Tier 3), the chunked approach also solves the memory problem. keeping a 30-minute audio file in memory for transcription will crash on lower-end devices. streaming chunks to the model keeps memory flat regardless of recording length.

the 100% offline approach is the right bet. users who care about privacy enough to choose Dictus over Google's built-in dictation will not accept any cloud fallback.

[m13v]

our on-device transcription with the chunked processing and vocabulary priming I mentioned: https://github.com/m13v/fazm/blob/main/Desktop/Sources/TranscriptionService.swift

macOS/Swift implementation but the whisper.cpp patterns (chunking, overlap, initial_prompt) are the same regardless of platform.