Speakr is a personal, self-hosted web application designed for transcribing audio recordings
Python
Commits on master
v0.9.0 docs Wave 2: thorough screenshot refresh + screenshots.md gallery update — replaces 22 existing screenshots and adds 8 new ones, all captured via Playwright MCP from spdev.murtaza.cc in dark mode at the resolutions matching each file's original aspect (1716x1270 for main app shots, 1880x1427 for the upload modal flow, 414x896 for mobile); refreshed: Main view, main-view-with-speakers, main-view-summary-tab, main-view-notes-tab, audio-sync-bubble-view, chat-interface, filters-sidebar-expanded, Inquire mode, recording-upload-modal, Share modal, edit recording modal, edit recording tags modal, edit-tags, edit-transcription-modal, summary reprocessing modal, full recording reprocessing, settings (account / custom prompts / shared transcripts / speakr management / about), tag-management-enhanced-badges, speaker-identification-voice-profiles, speakers-management-voice-profiles, Admin (dashboard / system settings / default prompts / stats / vector store); added: stats-tab.png, system-audio-help-(linux/macos/windows).png, recording-input-device-picker.png, mobile-(detail-view/transcript/summary/chat/stats/more-menu/upload-sheet).png — every shot shows the new v0.9.0 chrome (consolidated header, sidebar with section bands, sticky speaker pills with timestamp pills, design-system-unified modals, dock-right floating chat panel, mobile bottom nav with contextual icons in the chevron row, edge-to-edge content, mobile upload as bottom sheet, etc.); deleted two redundant duplicates (highlighted-transcript-following-simple-dark, audio-sync-simple-view) since their content was structurally identical to main-view-summary-tab in the new design; docs/screenshots.md gallery gains a new 'What's New in v0.9.0' section at the top with the new Stats tab, the three per-OS system-audio help tabs, the input device picker, and the mobile-first detail view shots (Summary / Transcript / Chat / Stats / More overflow / Upload bottom-sheet) so users browsing the gallery see the headline v0.9.0 changes first; Wave 2 admin webhooks doc reviewed and confirmed comprehensive — it already documents recording.transcription.started, events.extracted, recording.updated, HMAC signing, SSRF guard, retry policy, all env vars; recording-sessions.md similarly comprehensive — no Wave 2 doc updates needed beyond what already shipped with the original phase work Murtaza Nasir committed 5 months ago
f51821e v0.9.0 docs + README + screenshot refresh — Wave 1 of the planned doc sweep (end-user docs only; admin webhook deep-dive and screenshots.md gallery rebuild are deferred to a follow-up); (1) release_notes_v0.9.0.md created at repo root following the v0.8.x release-notes style with sections for System Audio & Multi-Input Recording, Stats Tab, Upload Modal Redesign, Mobile UI, Inquire Mode, Design System Unification, Polish & Bug Fixes, Backend & Infrastructure, Localization, Compatibility; (2) README.md updated — version badge bumped to 0.9.0, Latest Release section rewritten to point at v0.9.0 (with the v0.8.21-alpha security patch moved into Previous Releases), Key Features list expanded with system audio + multi-input recording + recording stats + webhooks + mobile-first detail view + Brazilian Portuguese; (3) docs/index.md banner switched from warning (v0.8.21-alpha security) to success (v0.9.0 feature release); (4) docs/features.md gained a new top-of-page Recording & Capture section with Multi-Platform System Audio Capture, Multi-Input Recording, and Recording Stats feature blocks; (5) docs/user-guide/recording.md got the biggest rewrite — System Audio section now carries a platform support matrix table (Chrome/Edge × Windows/macOS/Linux/etc.), three calmer recording buttons explained as primary/secondary hierarchy, per-OS setup pulled from the help modal (macOS BlackHole + Multi-Output Device, Windows Share-system-audio checkbox, Linux pavucontrol three-option workflow with the pactl module-virtual-source one-liner), Input Device Picker & Multi-Input Recording section, Disable echo cancellation toggle explanation, Privacy Notes for virtual audio devices with per-OS mitigations; Upload Modal section rewritten for progressive disclosure / chip summary / sticky footer / file preview with duration probe / last-used defaults auto-restore / mobile bottom-sheet with drag-to-dismiss; (6) first-steps.md updated — single adaptive top header description, new Right Rail Tabs section with Stats and Events callouts; (7) transcripts.md gained Sticky Speaker Pills note and a Recording Stats section explaining the headline metrics + per-speaker breakdown + silence row; (8) inquire-mode.md added a tip box explaining the ?upload=1 deep-link from + New Recording; (9) settings.md added a Display subsection for the new audio-player-position preference; (10) pwa.md gained a Mobile-First Detail View section covering the bottom nav / single header / contextual icons in chevron row / edge-to-edge content / sticky speaker pills / editor / audio player / upload sheet drag-to-dismiss / progress queue bottom-sheet; (11) screenshot refresh via Playwright MCP from spdev.murtaza.cc — Main view.png and Inquire mode.png replaced at their original 1716x1270 dimensions; recording-upload-modal.png replaced at 1880x1427 to match upload-interface.png aspect; brand new screenshots added at matching sizes for recording-input-device-picker, system-audio-help-(linux/macos/windows), stats-tab, mobile-detail-view (414x896), mobile-transcript (414x896 — also shows the new bottom-sheet processing queue above the player + nav) Murtaza Nasir committed 5 months ago
ab5ed60 Inquire mode: polish across header, sidebar, chat input and welcome state to match the rest of the app's design system — was largely using bespoke Tailwind utility chains instead of the .btn / .field / .empty / .surface-raised / .t-* primitives shared with the main app; (1) header: mobile sidebar toggle and user-menu button switched from bespoke 'p-2 rounded-lg hover:bg-tertiary' patterns to .btn--icon and .btn--ghost.btn--sm so they share the visual rhythm of the rest of the app's icon buttons and chrome buttons; user menu's identifier text stays hidden on small screens like the main app's header does; (2) sidebar (Filters): mobile close button uses .btn--icon, 'Clear all' switched to .btn--sm.btn--ghost (was 'text-xs px-3 py-1 bg-button text-button rounded-md'), Tag / Speaker / Date Range cards switched to a tighter rounded-md border padding-3 surface with t-body-strong section headings, search inputs use .field primitive with var(--fs-meta) for compact sizing, no-speakers state replaced with the design-system .empty primitive instead of bespoke centred chrome, Active Filters Summary card converted from a stack of bg-tertiary lines to a wrap of bg-accent pill-status chips so it reads as 'this is active right now'; (3) chat area: welcome state restructured with a circle icon badge + clear headline + supporting copy + a .surface-raised example questions card with chevron-right bullets so it reads as polished landing instead of a stacked text wall, chat message bubbles use the existing .message.user-message / .message.ai-message primitives so they inherit consistent prose styling, loading indicator uses .spinner primitive and progress step bars use inline :style for theme-aware colours instead of conditional Tailwind utilities; (4) chat input row: textarea uses .field primitive, send button uses .btn--primary.btn--icon.btn--lg matching the main app's chat send, hint row uses .t-meta type primitive; (5) toast container uses the shared modals/toast-container.html partial so inquire gets the same header-offset positioning (top: var(--app-header-height) + 16px) the main app uses Murtaza Nasir committed 5 months ago
e628454 Mobile detail: contextual icons move to the chevron row, panels go edge-to-edge, follow-player toggle stops looking bigger than its peers — three coupled changes that together fix the 'box within a box' and 'icons take a whole extra row' problems from the screenshot; (a) detail-view.html: the outer wrapper around the mobile panels dropped its p-4, so the per-tab panels now own their own edges instead of sitting inside a 16 px gutter inside another 16 px gutter; (b) mobile-header.html: a new contextual toolbar slot appears IN the chevron row, populated based on mobileTab — summary / notes show copy / download / edit (hidden during editing), transcript shows follow-player + view-mode + copy / download / edit, events shows calendar-export, chat shows clear-chat-when-not-empty; saves a row of vertical space across every panel; (c) each panel dropped its own top toolbar div (mobile-summary-panel, mobile-notes-panel, mobile-transcript-panel, mobile-chat-panel, mobile-events-panel) and tightened reading-mode padding to px-3 py-3 (was px-3 pb-3 with an additional outer p-4); editing-mode shells now use just pt-1 with no horizontal padding so the editor toolbar goes screen-edge to screen-edge for the largest possible typing area, with the sticky Cancel/Save footer providing its own px-3 py-2 for the buttons; chat panel's floating absolute clear button removed (it was floating over message content) since the same action is now in the header; (d) follow-player toggle (from the user's follow-up): when ON, only the icon COLOUR changes to text-accent — the previous bg-accent pill made the button visually heavier than its peers in the same toolbar; button dimensions stay identical to copy / download / edit regardless of active state, fixing the 'inconsistent box sizes' issue from the follow-up message Murtaza Nasir committed 5 months ago
ebd8127 Privacy notes polish — promote heading to a real h4 with a warning triangle icon, drop meta-explanation phrases that didn't add information, replace remaining em-dashes throughout the help modal with periods / commas / colons; (a) Privacy notes summary now uses fas fa-triangle-exclamation in warning-amber + an <h4 class='text-base font-semibold'> for the title so it reads as a prominent header instead of a t-body-strong run-in; chevron-down on the right indicates expandability; (b) dropped 'That's a deployment-config decision, not specific to these capture modes' from the Where-audio-ends-up paragraph, dropped 'No new privacy surface beyond what's already in any voice-recording web app' from the Microphone block, dropped 'The data path doesn't change; everything still goes to the same place' from the audio-processing block, dropped 'A couple of them have' tightened to 'A couple have', removed several other redundant phrasings; (c) all 22 em-dashes across the modal replaced with periods, commas, or colons as fits the prose; (d) double-dashes left untouched only where they're CSS BEM modifiers (modal-panel--lg, banner--success, etc.) which are code, not punctuation Murtaza Nasir committed 5 months ago
1c61501