v0.4

This release focuses on transcription capability and transcript annotation, and adds many features to the transcripts themselves as well as improves the methods that a user can use to run transcription.

Highlights

Transcript Sections

• Transcripts can now be broken in sections like chapters, to improve navigation and organization
• Sections can be added, removed, renamed, and moved easily in the transcript editing panel

Audio Effect Ranges

• Effects can now be applied to parts of the transcript, which will be rendered into the audio itself
• Effects are prerendered, and only the effected file is served, ensuring that the original file is not available to listeners
• NOTE: Applying affects generates a second .mp3 file containing the effects, which may double the storage impact of that project

Transcript muting

• Word ranges in the transcript can now be easily muted, which removes the audio from the effected file and strikes out the text
• Muting uses the effects pipeline, but is handled separately from the rest of the effects chain to make it easier to use
• Muted audio never appears in presentations, and you can choose whether they are muted or skipped during playback

Transcript notes

• Notes can be added to transcripts, including editor notes, stage notes, and speech notes
• Notes can be edited in situ, and can be shown or hidden in presentation mode at the editor's discretion

Project Version History

• Projects now automatically save snapshots as you work, and you can manually save named versions at any time
• Browse version history to open an older version as a new project, or revert the current project in place (fully undoable)

Off The Record

• Paragraphs can now be marked as "Off The Record" via right-click on the paragraph handle
• Off-the-record paragraphs are visually distinct in the editor and cannot be used in Live Quotes
• In presentation mode, off-the-record paragraphs are always shown as a placeholder — their text and audio are redacted server-side
• A prerendered audio_otr.mp3 is generated whenever OTR status changes, ensuring muted audio is served even if the raw file is accessed directly

Parallel Transcription

• Multiple projects can now be transcribed at the same time without blocking each other
• Each transcribing project displays a circular progress indicator in the sidebar so you can track all active jobs at a glance
• Switching away from a transcribing project no longer interrupts it — the job continues in the background and the transcript loads automatically when it finishes
• If you reload the page mid-transcription, the app reconnects to the in-progress job and resumes showing its progress
• Monthly usage quota checks now account for all in-flight jobs simultaneously, so parallel transcriptions cannot collectively overrun a plan's limits

Desktop App Overhaul

• The desktop version has been overhauled to support both CPU and GPU builds, allowing you to use on-device transcription without a cloud subscription
• Windows installers are now available for both CPU and GPU builds — download, run the installer, and launch from your desktop or Start Menu
• GPU builds detect your CUDA version at startup and show a clear error if the installed drivers are incompatible, with guidance on how to resolve it
• A dedicated Downloads page makes it easy to pick the right build for your system
• Local transcription uses faster-whisper + pyannote to deliver speaker-diarized transcripts entirely offline

Payment Integration

• It is now possible to pay for your subscriptions, and manage your payments using Stripe
• While the app is in Beta mode, no actual charges will be made, and a stand-in visa card may be used

---

Changelog

Added

• Product rebranded from Waveform Studio to SourceQuote across README, docs, UI, local executable name, local data folder, and Live Quote footer
• Public changelog pages: /changelog index and /changelog/ pages render release notes
• License page: /license renders the bundled AGPL license
• Installer download endpoint: /download/installer/// resolves and serves installer builds
• Admin dispatch system: admin UI and API to preview, count recipients, and send broadcast emails with changelog insertion support
• Admin changelog API: admins can fetch rendered changelog files for dispatch/email composition
• Live Quote embed improvements
• Embeds can now have names/titles
• Optional title display on embed
• "Copy link" button added to embed management
• Environment/config additions: HF_TOKEN_LOCAL for desktop binaries; SMTP_FROM_NAME for email sender display name
• Scheduled maintenance documentation: README now documents Windows scheduled tasks for analytics collection and usage rollover
• Presentation mode transcript search: search controls, match counts, next/previous navigation, and Ctrl+F-style keyboard behavior
• Email notification system expanded: HTML email templates for email-change alerts, admin dispatches, transcription-complete notifications, password reset links, and password-change notices
• Users can now change the email associated with their account
• Database now tracks page visits for all html-serving routes
• Admin analytics now display page visit tracking
• Page visit tracking updated to include all routes including api routes and error pages
• Updated database & admin page to save and display route methods and error status for tracking
• Transcript sections
• Sections are defined by their temporal position in the transcript
• Sections can be renamed in the transcript section break headers
• When sections are defined, the Sections panel appears for easy navigation
• When sections are defined, the section breaks are represented with markers in the waveform.
• Section breaks can be moved by dragging the header
• Sections can be viewed and browsed in presentation mode
• Effect ranges
• Effects can be applied to text selections and snap to word-level
• Effects ranges are stored by timecode, and apply to the audio itself
• When effects are added, a new modulated_audio.mp3 file is created in the project directory, and is spot-rendered any time an effect change is made.
• Muting is an effect, but is handled separately for a couple of reasons
  • Easy use, since muting is likely the most frequently utilized effect
  • Visual separation, as most people don't think of muting as an effect, and more a state
  • Makes skipping muted ranges simpler and more intuitive
• Retranscription is now allowed for any segment
• Off-the-record paragraphs
• Paragraphs can be marked or unmarked as "Off The Record" via right-click on the paragraph handle
• Off-the-record paragraphs are displayed with a hatched handle bar, dimmed italic text, and an "Off The Record" badge
• Selections that include an off-the-record paragraph cannot be used to generate a Live Quote
• In presentation mode, off-the-record paragraphs show a placeholder and their audio is skipped during playback
• Text and audio data for off-the-record paragraphs is redacted server-side — the presentation frontend never receives it
• The flag is stored as off_the_record: true on the paragraph in transcript.json and is fully undo/redo-able
• When OTR is applied or removed, audio_otr.mp3 is re-rendered in the background with OTR sections muted; presentation audio always serves this file when present, falling back to audio_effects.mp3 then audio.mp3
• OTR audio is re-rendered automatically whenever the effects base changes (effects added/removed)
• Project version history
• Auto-snapshots fire after 20 minutes of inactivity or every 50 edits
• Named versions can be saved manually with an optional label via the History button or context menu
• Versions can be opened as a new project (original version stack is untouched) or reverted in place (revert is undoable)
• Viewers can browse and open versions as new projects; editors and owners can save, revert, and delete
• Auto-version cap is configurable per user in account settings (default 50); named versions are uncapped
• Each version records who created it, ready for future multi-user attribution
• Parallel transcription
• Multiple projects can be transcribed simultaneously; each job runs independently in a background thread
• Active transcriptions are tracked server-side so progress survives page reloads — the client reconnects and resumes polling on load
• Circular progress indicator appears on the left of each transcribing project's sidebar entry
• Switching projects while a transcription is running no longer cancels it; the transcript reloads automatically on completion
• Monthly quota check includes in-flight reserved minutes so parallel jobs cannot collectively exceed a plan's limit
• Added 'email when done' option to the transcription dialog which will send an email to the user when transcription completes
• Added skip-to paragraph buttons in the waveform panel. Pressing Shift with arrow keys will activate them as well
• Desktop app overhaul
• CPU and GPU builds now available as standalone Windows installers (Inno Setup)
• Local transcription via faster-whisper + pyannote runs fully offline with speaker diarization
• GPU build checks CUDA version at startup and shows a user-friendly error if drivers are incompatible
• New Downloads page for choosing and downloading the right build
• Projects data directory moved from ~/Documents/WaveformStudio to ~/Documents/SourceQuote
• GPU and CPU PyInstaller specs split into install/local-gpu.spec and install/local-cpu.spec
• Password can now be reset in the settings
• Added a 'forgot password' password reset option
• Added tutorial slide show on open with settings option to disable
• Stripe integration
• Stripe is now used to facilitate payment for subscription charges
• Users will be converted to customers if they have no customer id
• Modifying subscriptions in the admin panel will update them on Stripe
• When in Beta mode, it will use the stripe sandbox and not make any real charges

Changed

• Live Quote embeds are now soft-deleted with a "deleted" tombstone instead of being removed; deleted embeds stop incrementing invoke count
• Transcription minute reservation: user usage now reserves and reconciles minutes for in-flight jobs, not just checks quota at start
• /api/fetch-page-title endpoint is now properly routed and tracked for hyperlink/title workflows
• Fixed laggy word-level highlight behavior in Live Quotes (This improvement will not be applied to previously generated Live Quotes.)
• Gave admins the ability to edit public files
• Fixed cron scripts calling the wrong venv
• Switched from using the native wavesurfer coloring system to applying color to it externally to allow for colored waveform regions
• Moved annotations from annotations.json to a key in transcript.json
• Fixed project names not saving
• Fixed bug report dialog not correctly capturing keypresses
• Uploaded MP3 files are now re-encoded through libmp3lame on ingest to guarantee an accurate Xing VBR seek table. Files with a linearised TOC (a known bug in certain ffmpeg/LAME configurations) caused audible seek drift on recordings longer than ~20 minutes.
• Reworked context menus to share common parent and handle menu stacking
• Added context menus to more UI elements
• Adjusted waveform controls layout and icons
• Switched editor to use "next segment" buttons instead of "skip 5s" buttons
• Skip-to paragraph and skip-to segment now use up/down arrow keys in addition to left/right for transcript/waveform direction discrepancies
• Keyboard shortcut legend is now collapsible
• App now opens with the editor divider centered
• Potential fix for stale audio on token expiration
• Ensured all currently active keyboard shortcuts are available in the shortcuts panel
• Switched account settings to be on a dedicated page instead of a drawer iframe