HAI HAI
Home ProjectsProfessional portfolio Contact

Local dictation and meeting notes for Windows

Notewll

Private Windows dictation, notes and meetings with a simpler workspace, clear word rules and always-visible voice analytics.

Windows 10 / 11Offline EnglishNo account needed
Download Windows preview - 0.11.0 ↗View source ↗Try demo
Notewll project preview

Try live voice notes

Speak. Watch your note appear.

Notewll turns speech into written notes. In the Windows app, use your shortcut to write in notes, browser text fields, editors and terminals.

This browser demo uses your browser's speech service. The Windows app runs standard English speech locally.

Choose Start dictation and speak. Or focus the button, hold Space, speak, and release. Words appear here as you speak; the note gets sentence capitalization and punctuation.

Ready. Start dictation and speak to write your note.

English dictation uses your browser's speech service, which may send audio to its provider. Browser support varies; Chrome or Edge is recommended for this demo. Sentence cleanup fixes basic casing and punctuation; review the words for recognition mistakes. This page does not type into other applications. Each session stops after one minute.

New in v0.11.0

A simpler Windows workspace

Use one Create menu for Dictate, Note, Meeting and Conversation. Review saved work in History, manage spellings and snippets in clean two-column tables, and keep speech and notes on your Windows PC.

One Create menu
Start Dictate, Note, Meeting or Conversation from a compact Home card. Try Notewll opens the local editor.
Analytics, always visible
See estimated time saved, words, sentences, active days and the weekly savings chart without expanding a section.
Clear words and snippets
Search inside centered fields, select rows in aligned two-column tables and save edits immediately. Dictionary fixes recognized spellings; Snippets expand short spoken phrases.

What it is for

Use one Create menu for Dictate, Note, Meeting and Conversation. Review saved work in History, manage spellings and snippets in clean two-column tables, and keep speech and notes on your Windows PC.

Capture thoughts with less navigation. Review recognized words, customize how speech becomes writing and return to saved conversations without a cloud account.

Speech to text for Windows
Dictate a draft in a focused text field, let supported repeated starts clean up, then review before sending.
A local note-taking workspace
Combine personal voice notes and meeting transcripts with original statements, audio and your own annotations.
From conversation to follow-through
Find supporting passages, jump to timestamps, confirm suggested tasks and keep local recovery snapshots.
Your voice, your orbit
Keep a playful floating ball with physics, skins, gravity choices, custom colors and local images.

Get started

Download Notewll-Setup-0.11.0.exe, finish recordings and close Notewll, then run the installer. Notes and preferences are preserved. On Home choose Create or open Try Notewll. Set text behavior on Style. Dictionary and Snippets save entries immediately; History contains your saved work. Additional languages require Optional features > Set up multilingual speech. This is an unsigned Windows preview.

Download Windows preview - 0.11.0 ↗View source ↗

Features and workflow

One Create menu
Start Dictate, Note, Meeting or Conversation from a compact Home card. Try Notewll opens the local editor.
Analytics, always visible
See estimated time saved, words, sentences, active days and the weekly savings chart without expanding a section.
Clear words and snippets
Search inside centered fields, select rows in aligned two-column tables and save edits immediately. Dictionary fixes recognized spellings; Snippets expand short spoken phrases.
Consistent controls
Roomier dropdowns, rounded highlights, selected-option marks and matching light/dark styling throughout the Windows app.
Local writing styles
Choose Raw text, Light cleanup or Modified text on Style. Standard, Chat, Email and Code formats and optional process-specific styles keep writing preferences together.
Corrections you confirm
Review a proposed short word correction before saving it. Originals are retained; corrections are not silently learned from another application.
Formatted snippets, by choice
Save bold, italic, bullets and URL text. Exact-trigger formatted paste is opt-in; targets may use the plain-text fallback and clipboard restoration is best effort.
Previewed voice edits
Use explicit local commands such as replace X with Y, uppercase and remove the last sentence, then review the result before applying it. This is not open-ended AI rewriting.
Local calendar reminders
Import individual timed ICS events and opt into reminders while the app runs. Recording always starts with your action. Recurring series and account synchronization are not supported.
Optional multilingual speech
Additional languages and auto-detection use a separately installed local Whisper base model. The standard bundled path stays offline English; multilingual quality needs real-audio acceptance.
Safer local storage
Atomic note saves, recovery copies and validated imports preserve committed content when a save or import fails. Startup avoids repeated history scans.
Raw text
Keep recognized text unchanged, including fillers and repetitions. Dictionary replacements and writing-style formatting are bypassed.
Modified text
Remove fillers and repeated starts with basic capitalization, punctuation and your selected writing preferences. No sentence-editing model is needed.
Find the moment, not just the file
Find & act connects optional local semantic retrieval to supporting transcript passages, source tracks and timestamps. It is evidence retrieval, not an open-ended chatbot. Model inference/resource acceptance for this preview remains pending.
Action items you review
Suggested tasks show their original statement, editable owner/date fields and a confirmation checkbox. They become tasks after review.
Hands-free development preview
Enroll your voice, select allowed apps and start explicitly. The app pauses on lock or microphone changes; Escape stops listening. Matching can make mistakes and physical desktop acceptance remains pending.
Preview, redact, share
Choose exactly what the PDF contains, exclude personal annotations and redact selected preview text. The original recording is unchanged; audio is neither included nor redacted.
A recovery plan for your notes
Opt into daily local snapshots while the app runs. Encrypted snapshots authenticate their contents and restore previews skip existing IDs. Encryption is tied to the Windows account/PC; ZIP supports migration.
One file. Ready for offline English speech.
The installer includes the Windows app, native speech runtime, English model and uninstaller. Standard installation needs no internet connection, Python, Docker, terminal commands or cloud account.
Your voice. Your orbit.
Drag the floating ball, release it and watch gravity, collisions and rolling friction bring it to rest. Click it to reveal an animated half-circle of readable controls.
  1. Download the 0.11.0 Windows preview installer and SHA256 checksums. Finish recordings and close Notewll before updating.
  2. Install and open Notewll; existing notes and preferences are preserved. Standard offline English speech is included.
  3. Choose Create on Home, or open Try Notewll to test the local editor. Test dictation in your chosen application.
  4. Set text behavior on Style. Manage recognized spellings in Dictionary and spoken expansions in Snippets. Entries save immediately.
  5. Use History to search saved work. Optional languages, semantic retrieval and voice matching have separate local setup and acceptance limits.

Architecture

Native C# WinForms on .NET Framework 4.8 captures microphone and WASAPI output audio. Bundled Vosk supplies standard offline English speech. Optional accurate speech, multilingual Whisper, retrieval and voice matching have separate local setup paths. Conservative writing rules preserve originals; notes save atomically.

Privacy and limitations

The Windows app processes speech and notes locally. Standard installation and recognition work offline. Meeting audio and original transcripts stay separate from editable personal notes. Optional recognition, semantic search and voice matching have their own setup and acceptance limits. Exported PDFs can be edited externally. Review recognized words before sharing.

Status

0.11.0 unsigned Windows x64 preview. 177 applicable automated tests passed, with 3 skipped and 3 excluded checks. The UI audit checked 210 captures with no layout findings. Physical microphone/call capture, cross-app insertion, multilingual accuracy, mixed-DPI hardware and Narrator remain separate desktop acceptance checks. No Microsoft Store submission or certification is claimed.

Version
0.11.0 unsigned Windows x64 preview
Headline
A simpler Windows workspace
Core
Offline English dictation, voice notes and meeting transcripts
Home
Compact Create menu and always-visible voice analytics
Dictionary and Snippets
Immediate saving, centered search and aligned two-column tables
Text modes
Raw text, Light cleanup and Modified text
Optional local features
Formatted snippets, voice-edit commands, multilingual speech and ICS reminders
Verification
177 applicable tests passed; 210 UI captures with no layout findings

Evidence

0.11.0 Windows application, installer and UI verification
0.11.0 Windows preview · reviewed 2026-10-29
177 applicable tests passed; 3 skipped and 3 excluded. 210 UI captures with no layout findings. These checks do not establish physical microphone/call capture, cross-app insertion, multilingual accuracy or Store certification.

0.9.0 installer, physics and customization checks
0.9.0 Windows preview · reviewed 2026-10-24
119 automated checks passed; four live hardware/desktop checks excluded. Clean isolated bundle extraction loaded the included offline speech engine. Ball checks cover all 171 appearance/gravity combinations and local-image rotation. Full interactive installation and physical desktop capture/insertion remain separate acceptance checks.

Previews — separate from stable releases

v0.11.0 · Preview

2026-10-29

A simpler Windows workspace: compact Create menu, always-visible analytics, History focused on saved work, clean Dictionary and Snippets tables, matching search fields and restyled dropdowns. Includes local writing controls, optional formatted snippets, previewed voice-edit commands, optional multilingual speech, local ICS reminders and safer atomic note storage. Unsigned preview; desktop capture/insertion and multilingual accuracy remain acceptance checks.

Download Notewll-Setup-0.11.0.exeDownload SHA256SUMS-0.11.0.txt

v0.10.0 · Preview

2026-10-24

# Notewll 0.10.0 — Speak naturally. Keep the words that matter. Local dictation, voice notes and meeting transcripts for Windows. ## The headline: two text modes Choose Raw text to keep recognized words unchanged, or Modified text to remove fillers and repeated starts with basic formatting. Original text remains available for review. ## Everything new in this preview - Raw text and Modified text are the only text modes. Raw text bypasses cleanup and writing rules; Modified text applies basic cleanup. - General repeated-prefix and abandoned-start cleanup, straight/curly apostrophe matching and two text modes. - Find & act across conversations, optional local semantic retrieval with source playback, extractive highlights and reviewable action items. - Timestamped live bookmarks: Ctrl+Alt+1/2/3 for Important, Question and Follow up. - Projects, tags and person filters; clear correction-rule scopes and undo through rule removal. - Experimental enrolled-voice hands-free mode with an app allowlist, visible state, lock/device-change pause and Escape to stop. - Exact-text PDF preview redaction. Personal notes can be excluded; source audio is not redacted or included. - Opt-in encrypted daily snapshots, authentication/tamper checks, restore previews and existing-ID protection. - Managed optional-feature storage information and install/remove actions. Unknown download/RAM measurements stay labeled unknown. - Current screenshots and an illustrative cleanup animation; the existing customizable physics ball remains. ## Downloads - **Notewll-Setup-0.10.0.exe**: full Windows x64 installer with standard offline English speech included. Run it to install or update, preserving notes/preferences. - **Notewll-Setup.exe**: the same installer under a convenient filename. - **Notewll-Small-Update.zip**: compact update files and instructions for an existing installation. Not a full installer. - **SHA256SUMS-0.10.0.txt**: checksums for those assets. ## Evidence and preview limits The text-mode update passed 164 automated checks, with 2 skipped checks and 2 excluded live microphone checks. Original text, saved settings and installer packaging were checked. Physical microphone capture and typing into target applications still require desktop acceptance. This is an unsigned Windows preview. [Product, demo and screenshots](https://myhai.org/projects/dotflow/) · [User guide](https://github.com/sudheer-050/notewll/blob/main/USER_GUIDE.md) · [Source](https://github.com/sudheer-050/notewll)

Download Notewll-Setup-0.10.0.exeDownload Notewll-Small-Update.zipDownload SHA256SUMS-0.10.0.txt

v0.8.0 · Preview

2026-10-23

# Notewll 0.8.0 - faster local dictation - Shortcut dictation now recognizes speech every 250 ms while you speak, including when live insertion is disabled or held modifier keys delay typing. - Silence, quick shortcut taps and empty recognition results return quietly to idle without warning popups, inserted text or clipboard changes. - Playground preview no longer clears a computed result when an empty optional result is supplied. - Existing notes, recordings, shortcuts and preferences are retained. Installer, application metadata and About now report 0.8.0. Verification: 91 application, installer, speech-worker and regression checks passed; four live microphone/keyboard checks were excluded. Direct empty-result/cancellation checks passed too. The full installer embeds the exact verified 0.8.0 application. A cleanup-preview failure found during packaging was fixed and both preview checks now pass. Real microphone capture returns MMSYSERR 1 in this managed session, and keyboard injection is unavailable. Physical shortcut timing, cross-app insertion and recognition accuracy still need desktop acceptance. The installer is unsigned and the certificate-verified automatic updater remains unconfigured; install this preview manually from GitHub Releases. See verification/STREAMING_DICTATION.md for the measured scope and remaining checks.

Download Notewll-Setup-0.8.0.exeDownload SHA256SUMS.txt

v0.7.0 · Preview

2026-10-22

# Notewll 0.7.0 Windows x64, English-only, offline speech after initial setup download. - PDF-only conversation export with readable scope previews and explicit participant-label wording. Original words remain read-only; external PDF copies are not tamper-proof. - Responsive export, playback, storage and backup interfaces; separate retention and deleted-record recovery. - Saved-record review separated from new recording controls; actual saved audio sources only, device/activity feedback and grouped actions. - Apply/Cancel for device changes; permanent toolbar hiding and snooze respected across note actions. - Compact Home, contextual recovery, consistent Notewll branding, improved themed tabs/dropdowns, searchable word/snippet forms and explicit correction saving. - Standard, Chat, Email and conservative Code styles; user-controlled Practice and guided first-use checks. Verification: 85 automated checks passed, four live checks excluded. Isolated UI checks preserve original words and preferences. See FIX_REPORT_0.7.0.md for all 38 review findings and remaining acceptance work. The installer is unsigned because no code-signing certificate was provided. Published as an unsigned preview. Installation and live desktop acceptance have not been verified in this managed session. Real microphone/call capture, browser/terminal injection, mixed-DPI pointer behavior, Narrator and consented speech-accuracy comparison require a normal desktop session. The managed session denies those operations. No macOS/mobile/cloud sync release is included. October 22, 2026 update: renamed the application to Notewll across the UI, setup, shortcuts, exports and documentation. Dictation uses direct typing without reading the target field or an automatic paste retry. Existing LocalDictation data and installation identifiers are retained for upgrades. Shortcut preferences are preserved. The Windows preview installer is unsigned; real cross-app insertion still needs desktop acceptance.

Download Notewll-Setup-0.7.0.exeDownload SHA256SUMS.txt

v0.6.1 · Preview

2026-10-21

# Notewll / Local Dictation 0.6.1 release notes Windows x64 preview. This historical installer appears as Local Dictation. The project and current application are now named Notewll. - Safer native speech-engine shutdown, in-memory dictation retry and clipboard restoration. - Compact Home actions, recent work, persistent session issues, grouped preferences and first-run guidance. - Light/Dark/System themes, live system-theme refresh, high-contrast button/list colors, reduced motion and optional voice animation. - Edge-docked hover toolbar, one-time drag hint and keyboard action menu. - Continuous per-source meeting recognition, separate meters, pause/resume and explicit device-change recovery. - Content search, sort by date/title/duration, tags, editable speaker labels that preserve original words, and timestamp playback. - Local source-quote overview, possible decisions/follow-ups and keyword answers with timestamps. No new AI model, cloud service or invented speaker attribution. - Previewed export scopes, remembered format, searchable/tagged Unicode PDF, Word/Excel convenience protections and CSV. - Recently deleted recovery, opt-in retention, size breakdown and atomic note backup. Imports validate all metadata before moving records and never overwrite existing IDs. - Opt-in startup update checks and manual hash/publisher verification. Unsigned installers are rejected by the updater. First installation needs internet for the native runtime and small English model; speech afterwards is offline. No Docker, Python installation or cloud account is needed. The setup executable is a bootstrapper, not the whole installed runtime. This installer is unsigned: no Windows code-signing certificate is configured. It is published as a preview, not a signed automatic update. Document editing restrictions are removable and are not tamper-proof evidence. Verification: 79 application checks passed, four live checks deselected. Real installed Vosk passed 16-stream concurrent shutdown and continuous both-source silence tests. Synthetic silence, hum, clicks and quiet noise produced no words, with and without the quiet gate. Windows media playback seek/speed/pause/close passed on a fixture. Remaining acceptance limits: this managed session denies microphone/call capture and keyboard injection, and blocks Windows synthetic speech generation. Human speech accuracy, real calls, terminal/browser insertion, audible playback, mixed-DPI monitors and Narrator still need a normal desktop session. No direct Wispr accuracy comparison was performed. Source-quote review is lexical, not a semantic AI summary or generative Q&A. PDF structure tags are present, without a PDF/UA certification claim. Windows English-only remains the platform scope; no mobile/macOS/cloud sync build is included.

Download LocalDictation-Setup-0.6.1.exeDownload SHA256SUMS.txt

Questions

Which text mode should I choose?

Choose Raw text to keep recognized words unchanged. Light cleanup removes fillers and simple repeats; Modified text also handles supported spoken restarts and writing rules. Set these on Style. Originals remain available.

How do Dictionary and Snippets differ?

Dictionary corrects recurring recognized words and names. Snippets expand a short spoken trigger into longer text. Both use searchable two-column tables and save entries as soon as you save their dialog.

Does it work offline?

The included standard English speech engine and installation work offline. Optional feature setup needs downloads; local inference can run on installed model files. The browser demo is separate and may use online browser speech services.

Can I search meetings and jump to source audio?

Find & act implements passage retrieval and timestamp/source playback. Semantic search needs optional installation and its model-inference/resource acceptance remains pending in this preview.

Are action items created automatically?

Candidate tasks remain suggestions until you review owner/date/text and confirm them. They link to supporting statements.

Are my private annotations included in exports?

You choose Conversation only, Personal notes only or Full record, preview the text, and redact selections before saving a PDF. Audio is not included or redacted.

Is hands-free safe to leave running everywhere?

Choose an explicit allowed-app list and start it manually. Voice matching is experimental and can make mistakes. Escape pauses it, and lock/device changes pause listening. Desktop acceptance remains pending.

What happens to notes when I update?

The installer preserves local notes, recordings and preferences. Encrypted snapshots require the same Windows account/PC; keep ZIP backups for migration.

Is the installer signed?

This is an unsigned Windows preview. Versioned installer files and SHA256 checksums are published together on GitHub Releases.

Explore all projects →