Opportunities › Offline AI voice-to-text for developers › Proposed app
Hushdraft
Private voice dictation for developers that works offline in your browser
- Score
- 23
- confidence 45
- Evidence
- 3
- items on the page
- Rivals
- 2
- named in sources
- First version
- 35 d
- estimate
- Type
- Desktop app
- PWA: runs in the browser, installs on Mac and Windows
- Market since
- 1997
- Mature market · AI estimate
ScreensConcept
Desktop app: Developers use this at their desk while coding, with global hotkey capture, code-aware modes, and custom vocabulary management tied to their development workflow. Sample screens on a computer, in the proposed brand. Their numbers are sample data.
Dictate
Audio stays on deviceListening
About this appAI proposal
Hushdraft turns speech into text on your own machine. A speech model is downloaded once, cached by the app, and then runs locally in the browser or installed PWA. No audio leaves the device, no account is needed to dictate, and no API key is involved.
Made for developers, founders and AI power users on macOS and Windows who dictate prompts, commit messages, notes and code comments. Spoken text can be shaped by developer modes: it keeps identifiers in camelCase, expands spoken symbols, and strips filler words.
The difference from cloud dictation tools is simple: it keeps working on a plane, behind a firewall or on a bad connection, and your voice stays private. Custom vocabulary lets you teach it your library names and team jargon.
- Tagline
- Say it out loud. It never leaves your machine.
- Voice
- CalmDirectPrivate
- Why this name
- Hush says quiet and private; draft is the first version of a thought or a line of code. Two plain words, easy to say.
In the first version
Members say which features they would use. It tells PWAS what people want; it changes no score.
Fully offline after first setup
The model is cached on your device. Dictation works with no connection, so there is no network latency and no server to trust.
No member yet
Code-aware dictation modes
Say "open paren" or "snake case user id" and get ( and user_id. Switch between prose, code comment and commit message modes.
No member yet
Custom vocabulary
Add framework names, function names and acronyms so they are spelled the way you write them, not the way they sound.
No member yet
Global hotkey capture
Hold a key to talk, release to get text on the clipboard, ready to paste into your editor, terminal or chat.
No member yet
Brand identityAI proposal
Check the name against trademark registers before using it.
Palette
#2F5D50Pine · primary#F6F1E7Paper · surface#1F2A24Ink · text#D9822BAmber · accent#3B6EA5Slate blue · data
Typography
Aa
Hushdraft
ABCDEFGHIJKLMNOPQRSTUVWXYZ 0123456789
Say it out loud. It never leaves your machine.
PricingAI proposal
The material gives no competitor prices. These tiers are a starting point to test, based on the monetization idea of a one-time licence per developer with a team tier.
Free
Free
- Tiny model
- Prose mode
- Local history for 7 days
Solo
$29 once
- All model sizes
- Code and commit modes
- Custom vocabulary
- Unlimited local history
Team
$6 a month
- Everything in Solo
- Shared vocabulary packs
- Seat management
- Priority support
Would you pay Team · $6 a month?
Your answer tells PWAS which prices people accept. It changes no score.
No answers yet.
What it could be worthEstimate
What a software business making this much sells for today, from real deals. Pick a monthly revenue.
If it makes
Worth about
$194,000
Sale range $132,000 to $263,000
$120,000 a year in revenue
From 87 software businesses listed and sold on Empire Flippers in the last year, revenue vetted by the broker: they sell for about 19× their monthly revenue, most between 13× and 26×.
Measured on profit, the middle deal is 32× monthly net profit: about 2.7 years of profit.
An estimate, not an offer or financial advice. A real price depends on profit, growth, churn, how much of the work is the owner's, and the buyer.
How it would growAI proposal
How the product would reach new customers through its own use, and how the first ones would be found. A plan to test, not a result.
- Built-in loop
- No loop in the product. Dictation output is pasted into the user's own tools, so nothing carries the product's name to others. Growth would come from content, developer communities and word of mouth.
- Do they become customers?
- Reach comes only through sales and content. Test whether privacy-focused developers on forums convert from a free tier to a paid licence at a useful rate.
- First customers
- Independent developers and founders on macOS or Windows who already dictate prompts and notes and care about privacy, working alone or in teams under ten.
- Where to reach them
- GitHub and open-source forums: Post a demo showing dictation with the network disabled, and link to a free tier.
- Developer communities: Share in Hacker News, Reddit programming groups and AI tooling Discords with a privacy comparison.
- Short demo videos: Show code dictation and a commit message flow in under a minute.
- First milestone
- Fifty free installs that dictate offline weekly, and ten paying Solo licences from developers within two months.
Why people already pay for this
- VerifiedGitHub repository gained 728 stars in 460 days
- Public claimThere is market demand for offline, privacy-first voice-to-text tools targeting developers and power users.
- Public claimExisting solutions (Whisper Flow, Superwhisper) are positioned as alternatives, suggesting developers are actively seeking better options.
Tech stack
Every proposal builds on the same PWA base, installable from the browser on any phone or computer. Dashed boxes are what this app adds.
- AppNext.js and TypeScript
- PWAManifest and service worker
- UITailwind CSS
- Data and authSupabase Postgres
- PaymentsStripe
- EmailResend
- HostingVercel
- This app · Speech recognitionOpen-source Whisper-class model in WebAssembly/WebGPURuns transcription inside the browser so audio never leaves the device.
- This app · Local storageIndexedDB with Cache Storage for modelsKeeps model files and history on the device for offline use.
RisksAI inference
- TechnicalIn-browser models may be slower or less accurate than native apps, and need WebGPU support and several hundred MB of storage. Older laptops may struggle.
- PlatformA PWA cannot easily offer a system-wide hotkey or type into other apps. Output may be limited to clipboard paste, which could disappoint users of native tools.
- CommercialExisting tools already serve this audience, and open-source models make copies easy. Willingness to pay for an offline tool is unproven.
- PlatformBrowsers may evict cached model files under storage pressure, forcing a re-download that needs a connection.
- AICode dictation is error-prone: homophones, symbols and identifiers may be mis-transcribed, which erodes trust.
Everything marked AI proposal or AI inference was written by a model from the evidence on the opportunity page. The proposed name must be checked against trademark registers before it is used.