Updates

Publish a folder so anyone with the link can browse shared transcripts, text documents, and OCR documents—including nested folders.

• Visitors can browse published folders in a read-only view.
• Use the same folder link for owners, invited collaborators, and public viewers.
• Existing shared-folder links continue to work.

Follow your recording with synchronized text in Listen Mode, then export a vertical video clip with audio and highlighted words.

• Listen Mode highlights words as the recording plays.
• Choose a clip up to 2½ minutes long and preview it before exporting.
• Download a vertical MP4 with the original audio and synchronized transcript text.

Max users can import audio from Google Drive folders, review files before starting, and track transcription progress.

• Connect Google Drive and choose folders for a one-time audio import.
• Review files before starting and monitor import progress.
• Nested folders are reflected in your document library.

Hybrid transcripts render more efficiently, playback highlighting is more reliable, and editing avoids unexpected cursor jumps.

• Follow playback with more consistent word highlighting.
• Edit transcript text without unexpected caret jumps.

Choose Hybrid, Hebrew letters, or English transliteration as your default for Hebrew words in primarily English transcripts. Override it for individual uploads whenever needed.

• Set your preference in account customization.
• New uploads use your saved preference while keeping the per-transcription setting editable.

Paste a direct audio download link to submit a recording for transcription.

• Submit a supported direct audio link without downloading and re-uploading the recording yourself.

Choose which transformation results to share, keep others private, and send document and folder invitations by email.

• Transformation results are private by default until explicitly shared.
• Control which saved transformations are shared with a document or folder.
• Recipients can leave directly shared documents or folders.

Translations remain available as documents change, update affected sections incrementally, and preserve images in OCR translations.

• Previously translated content stays available when the source is edited.
• Update affected sections without regenerating the entire translation.
• Images are preserved in OCR translations.

Use document status filters and folder bulk actions to organize larger libraries with fewer clicks.

• Filter documents by status to find the items you need.
• Select multiple folders to pin, unpin, or move them together, or delete empty folders.

Invite others to access shared folders and work together on your documents.

• Share a folder instead of sharing each document individually.
• Choose access permissions for the people you invite.

Keep related documents together in folders and pin important items for quick access.

• Organize your document library into folders.
• Pin frequently used documents and folders so they are easier to find.

Translations preserve document structure and formatting more faithfully, with better alignment between original and translated transcript paragraphs.

• OCR translations preserve tables, columns, and inline formatting.
• Transcript translations keep original and translated paragraphs aligned for easier comparison.

Customers can submit Sofer.Ai transcripts and documents for professional editing by Machon Aleh Zayis.

• Choose an editing level, select one or more documents, and see exact request pricing before checkout.
• Submitted versions are captured at checkout, so customers can continue editing their originals without changing the editing request.

Large OCR jobs now preserve completed pages and resume missing work after an interruption instead of starting over.

• Large PDFs are processed in smaller chunks with targeted retries for failed pages.
• Page counts and OCR credits are reconciled to the authoritative PDF page count.

OCR now preserves complex document structure more accurately, especially in multi-column Hebrew pages.

• Headings, columns, marginalia, tables, footnotes, alignment, and reading order are reconstructed more faithfully.
• Source images, OCR text, and translations use synchronized highlighting for easier review.
• Pages with clearly corrupted output are retried automatically.

Transformation results now leave more room for reading on mobile while keeping copy and download actions close at hand.

• Word, PDF, and text download formats are grouped into a compact menu.
• The transformation viewer adapts to the visible mobile browser area for a more comfortable reading experience.

Transcript playback now seeks more accurately in long recordings and stays aligned with the audio timeline.

• New transcriptions use one consistent audio file for transcription, playback, and downloads.
• Transcript text and audio controls load faster and more consistently.

Transcript edit mode now includes find and replace controls for locating matches, stepping through them, and replacing one or all occurrences.

• Use previous and next controls to review each match before replacing it.
• Whole-word and match-case options help narrow searches, with Replace and Replace all actions available in the editor.

OCR and transcript search now match Hebrew text whether or not the search term includes nekudot or cantillation marks.

• Search results and highlights handle menukad and non-menukad Hebrew more consistently.
• The same matching behavior now applies across OCR, transcript, and text-document search paths.

Transformation outputs can now be downloaded in bulk from export flows.

• Bulk downloads support document and transformation export targets.
• Exports can include outline headings when requested, generating missing outlines when possible.

Word downloads now preserve complex formatting more consistently, and canceling transcript edits reliably restores the last saved version.

• Headings, inline formatting, tables, blockquotes, code, and nested lists carry into Word exports more accurately.
• Cancel immediately discards unsaved transcript changes, including Hybrid view edits.

Translate transcripts and OCR outputs while keeping the original text beside the translation for easier review.

• Review the original and translated text side by side instead of switching views.
• Set your own translation style in the advanced settings

Download OCR results as PDFs directly from the OCR document view.

• Export OCR text into a PDF without leaving the OCR document.
• Save or share OCR output in a cleaner document format.

Download your documents in bulk.

• Export documents as PDF, Word, or plain text.
• Select multiple documents—or your entire library—at once.

Search across transcripts, documents, and OCR files from any page containing text.

• Search using natural language with Maishiv AI or perform traditional keyword searches.
• Receive answers with direct quotes from your content.Bring direct quotes from text
• Click any result to jump to the original text with highlighted context.

Build your Maishiv knowledge base faster by importing transcripts, documents, OCR files, and entire folders of Word documents in just a few clicks.

• Add all existing transcripts, documents, and OCR documents to a knowledge base at once.
• Filter by content type and choose exactly which items to include.
• Upload entire folders of Word documents when creating new knowledge-base content.
• Faster setup for large knowledge bases with fewer manual uploads and selections.

Usage limits reset on the first day of each month at 00:00 UTC.

• Usage is tracked within a monthly usage period that ends at 00:00 UTC on the first day of each month.
• All usage counters reset automatically at that time.
• The reset schedule is based on UTC and is the same for all users regardless of their local time zone.

API transcriptions that use multiple speakers or automatic speaker detection now add $0.30 per hour.

• Passing num_speakers greater than 1 adds the multi-speaker transcription charge.
• Using auto_detect_speakers also adds the same $0.30/hour charge.

Free plan accounts can create up to 5 transcriptions per calendar month.

• The limit resets at the start of each calendar month.
• Failed, cancelled, and insufficient-funds transcriptions do not count toward the monthly total.

Sofer.Ai now has one place for product updates, feature launches, and rule changes.

• The Updates nav item opens a timeline that is visible to signed-in and signed-out visitors.
• Each entry supports a type, summary, detail bullets, and optional labels.

Create a batch manifest directly from a podcast RSS feed. Users provide an RSS URL, and the API extracts podcast episode audio links into a standard batch manifest that can be used for batch transcription.

French is now available as a primary transcription language.

Important service notices can now appear directly in Sofer.Ai when an outage or degraded provider affects transcription.

• Status messages can be updated quickly without waiting on a new app release.

MyTAT and TorahAnytime short links are normalized correctly when importing audio by link.

Plan access and usage limits for transformations and automations now show clearer messaging.

• Transformation access is enforced by plan.
• Automation limit messages now explain what is available for the current plan.

Previously generated summaries remain accessible across plans.

Transcript and audio downloads now work more consistently across browsers and source sites.

API users have clearer documentation for looking up batch transcription results by client item ID.

Word document exports now preserve Hebrew word ordering more accurately.

Outlines can now be inserted directly into the main body of a transcript from the outline panel.

CSV exports now use encoding that preserves text more reliably when opened in Excel.

Batch transcriptions are less likely to get stuck during finalization.

Batch transcriptions can now be automatically tagged into user set categories.

Accounts with many transcripts should see the documents list load more reliably.

Browser audio uploads are more reliable, especially for formats that need client-side preparation.

OCR jobs in progress are less likely to show a false failure while the system is still retrying.

Word downloads now preserve transcript and summary formatting more consistently.

Speaker labeling is now more accurate and reliable across a wider range of audio recordings.

Bulk batch manifest limits were raised, and pending balance holds are handled more accurately.

Transcription can now automatically detect speaker count when creating a new transcript.

Batch transcription requests now validate speaker settings and item IDs more consistently.

Transcript pages now include an MP3 download option.

Audio files that cannot be decoded now fail more cleanly with better handling.

Users can add OCR usage when they need more OCR capacity.

Kol Halashon is now included in supported audio link sources.

Transcription processing is more reliable under heavier parallel workloads.

Several transcript download paths were made more reliable.

Past Maishiv question titles now load more reliably.

Batch transcription status updates were made less likely to bottleneck while processing larger jobs.

Transcription uploads now accept WMA audio files in the browser upload flow.

Speaker labels in transcripts now behave more consistently when transcripts are generated and edited.

Maishiv answers now follow the expected language behavior more reliably.

Maishiv knowledge-base loading and search startup were improved for accounts with existing documents.

• Users can begin searching sooner while supporting document data continues loading.
• Large knowledge bases are handled more efficiently.

OCR documents gained stronger editing and download support from the document viewer.

Batch API documentation and manifest handling were cleaned up so batch transcription requests are easier to structure correctly.

Batch uploads now preserve batch titles and file identifiers more reliably.

Maishiv limits were raised alongside the updated Maishiv experience.

Maishiv was rebuilt with a fuller question-and-answer experience, knowledge-base tools, ability to hear original audio, and and improved document selection

The billing portal now exposes invoice history, including non-subscription purchases like API balance top-ups.

Transcription timeout and failure tracking were improved so stuck jobs are handled more cleanly.

Batch transcription errors are tracked more clearly, improving recovery and status reporting for failed batch items.

Batch transcription file downloads were made more reliable.

Transcripts can now be categorized automatically and managed from the categories workflow.

Large-file upload handling and error messages were improved for transcription and OCR uploads.

OCR document pages gained Word and PDF download options directly from the viewer.

OCR gained an enhanced mode for higher-quality extraction when users need a stronger OCR pass.

Sofer.Ai added OCR document extraction with upload, processing progress, document viewing, search, and usage tracking.

OCR uploads were refactored to support larger files more reliably with direct upload and clearer validation.

Transformations now handle maximum content size limits more predictably.

New document and transformation setup screens were polished for a smoother workflow.

Checkout and upgrade flows were improved for smaller screens.

The Help page added tutorial content for transformations, with a banner pointing users to the new guide.