# Torah and Chassidus Translate - NotebookLM Video Source Pack

Use this file as a factual source in Google's Gemini Notebook (formerly NotebookLM). Pair it with the complete user guide and the screenshot files. It is intentionally repetitive about claims and guardrails so generated video does not drift into unsupported marketing language.

## Product truth

Torah and Chassidus Translate is a source-centered study and translation workspace for photographed or scanned Hebrew and Yiddish texts. It keeps the source page, editable OCR, AI translation history, page identity, references, learning tools, exports, and backup controls together.

The central product promise is not “AI knows the text.” It is: **the learner can keep the source visible, produce a structured first-pass translation, review it, and continue learning from the same project.**

Primary audiences:

- A learner opening a difficult source for the first time.
- A teacher preparing a bilingual handout or guided lesson.
- A translator comparing models, templates, terminology, and prior versions.
- A researcher organizing scanned pages, references, tags, notes, and exports.
- A library or shared-device administrator supporting supervised access.

## Demonstrated workflow

The documentation walkthrough is a real authenticated test, not a mockup. Version 1.0.6 (98) imported an original Hebrew learning sample, saved it as a project page, authenticated with a temporary Google Gemini key, requested a translation from `gemini-3.1-flash-lite`, persisted the result, and displayed source and translation side by side.

Suggested visual sequence:

1. `ipad-03-library-project.png` - the named project in the library.
2. `ipad-04-project-overview.png` - page, status, tags, and run configuration.
3. `ipad-05-parallel-translation.png` - the strongest product shot: Hebrew page on the left, bilingual structured result on the right.
4. `ipad-06-study-with-tutor.png` - explicit consent before project-grounded cloud tutoring.
5. `ipad-02-gemini-ready.png` - provider status and masked credential.

Do not animate or reconstruct a visible API key. Crop or blur any accidental credential-like text. The supplied provider screenshot is already masked.

## Feature map

### Capture and preparation

- Camera, image, file, and PDF import.
- Crop, rotation, page preview, and edited-image retention.
- OCR assistance with editable source text.
- Structured printed-page and volume identity rules.

### Translation and review

- Bring-your-own-provider-key configuration.
- Provider, model, template, language, input, and quality selection.
- One-page, selected-page, and whole-project translation.
- Side-by-side image, source, and translation.
- Translation cache and history for comparison and recovery.
- Source and footnote inspection.

### Learning

- Word Lens for terms in context.
- Reader Mode and text to speech.
- Beginner, advanced, and scholar summaries.
- Tags, notes, and review items.
- Typed and live voice tutoring grounded in the open project or page.
- Consent, transcript, and provider-cost controls.

### Output and durability

- Reader and Study Reader PDF.
- Markdown, DOCX, ZIP, XLIFF, TMX, CSV, and comparison outputs where supported.
- Content-specific export naming.
- Backup, recovery, and shared-device controls.
- Downloadable local resources and local-model options for compatible tasks.

## Educational story

The app helps by shortening the distance between encountering a difficult page and asking a useful question about it. The source remains visible, so the learner can compare instead of passively accepting generated prose. Project context lets the tutor explain the open page rather than answer as a generic chatbot. Notes, review items, tags, summaries, and exports turn a single AI response into a repeatable study process.

A good tutorial should demonstrate this loop:

**See the source -> create a first pass -> verify a phrase -> ask a focused question -> save what matters -> return for review.**

## Privacy and cost truth

- Provider credentials use operating-system secure storage.
- Cloud requests send the required prompt, page excerpts, images, or audio to the selected provider.
- Bring-your-own-key usage is billed and governed by that provider.
- App cost displays are estimates; provider invoices are authoritative.
- Voice use can contain separate text and audio modalities and may cost more than typed chat.
- Local models reduce cloud transfer for compatible tasks but consume device resources and require model downloads.
- Saved transcripts may be included in configured backup or device sync. Users can disable transcript persistence.

## Claims that are allowed

- “Keep the original page beside the translation.”
- “Translate and study scanned Hebrew and Yiddish source pages.”
- “Choose your provider, model, and learning workflow.”
- “Ask questions grounded in the project you are reading.”
- “Review AI output against the source.”
- “Export a reading or study copy with a content-specific filename.”

## Claims that are forbidden

- Do not say translation is perfect, authoritative, guaranteed, rabbinically approved, or free from hallucinations.
- Do not say all processing is local or private; cloud features send data to the chosen provider.
- Do not promise zero cost, unlimited use, or a fixed provider price.
- Do not imply the app owns or includes every book shown by a user.
- Do not invent App Store ratings, user counts, testimonials, endorsements, or performance percentages.
- Do not call generated citations verified unless the user has inspected them.
- Do not show a real API key, private scan, user transcript, or copyrighted page without permission.

## 60-second advertising outline

**0-6 seconds - The difficulty**
Visual: close crop of the original Hebrew sample.
Narration: “A meaningful page should not become inaccessible because the language or print is difficult.”

**6-16 seconds - Bring the page in**
Visual: project library, then project overview.
Narration: “Create a project, import a scan or PDF, and keep every page organized with its source.”

**16-32 seconds - The proof**
Visual: slow pan across the parallel translation screenshot.
Narration: “Choose your AI provider and translation setup. Then compare the original, source text, and structured translation side by side.”

**32-44 seconds - Learn, not just translate**
Visual: highlight terms, references, tutor icon, and tutoring consent.
Narration: “Ask for a simpler explanation, inspect a reference, save a note, or turn a passage into review.”

**44-53 seconds - Control**
Visual: masked Gemini readiness and cost controls.
Narration: “Bring your own provider key, see readiness and usage estimates, or use compatible local resources.”

**53-60 seconds - Close**
Visual: source and translation return to full frame.
Narration: “Torah and Chassidus Translate. Keep the source in view, and make the next question possible.”

On-screen footer: “AI output requires review. Provider charges and terms may apply.”

## Five-minute tutorial outline

1. Explain the source-centered goal and the limits of AI output.
2. Create a clearly named project.
3. Show provider setup without revealing a key.
4. Import, crop, and check the full page.
5. Explain OCR text and page-versus-volume identity.
6. Translate one page with a selected provider, model, and template.
7. Review parallel bilingual output against the image.
8. Demonstrate one learning action: a term, reference, note, review item, or focused tutor question.
9. Explain cloud consent, transcripts, token/audio cost estimates, and local alternatives.
10. Export a content-specific document and configure backup.

## Visual direction

Use a calm editorial style: warm paper, deep green, restrained mint highlights, legible black type, generous whitespace, and subtle motion. Prefer slow crops, pointer highlights, and side-by-side comparisons over floating device mockups. Preserve Hebrew right-to-left direction. Never mirror the source page. Keep text large enough for an iPad screenshot to remain readable in a 16:9 video.

## Pronunciation and naming notes

- Product name: “Torah and Chassidus Translate.”
- Chassidus: use the pronunciation familiar to the intended community; avoid changing the spelling on screen.
- Gemini Notebook: Google's current name for the product formerly called NotebookLM. It may still be recognized publicly as NotebookLM.
- OCR: say “O-C-R” on first use, then “text recognition” for a general audience.

## Recommended Gemini Notebook production workflow

1. Create one notebook for the launch campaign rather than mixing product documentation with unrelated research.
2. Add the complete user guide, this source pack, the PDF source pack, and the five authenticated PNG screenshots. Gemini Notebook currently accepts Markdown, PDF, and common image formats.
3. Keep the source names descriptive. When more than five sources are present, use labels such as `Product truth`, `Screenshots`, `Tutorial`, and `Campaign`.
4. Select only the sources needed for the artifact. Use the complete guide and screenshots for a tutorial; add the campaign source pack for advertising.
5. In Studio, choose Video Overview and customize format, output language, visual style where supported, and the steering prompt.
6. Use **Explainer** for the product tutorial. Use **Short** for an approximately 60-second English video where available. Use **Cinematic** only for a restrained English brand film and only when its age/account requirements are satisfied.
7. Let generation continue in the background. Google notes that a Video Overview can take more than 30 minutes.
8. Download the result and review every interface label, product claim, Hebrew crop, narration line, and disclaimer. AI-generated video can contain inaccuracies and audio glitches.
9. Run the quality-control prompt in `notebooklm-video-prompts.md`, correct the steering prompt or source selection, and regenerate when necessary.
10. Treat generated video as a draft. Add final captions, branding, accessibility checks, legal review, and platform-specific end cards in a video editor before publishing.

Current Google documentation says an uploaded source can contain up to 500,000 words or 200 MB, and the free tier supports up to 50 sources. Plan and generation limits vary by account and are subject to change, so verify the account's current limits before scheduling a campaign.

## Official Gemini Notebook references

- Video Overview formats, customization, generation, sharing, and AI accuracy warning: https://support.google.com/gemininotebook/answer/16454555?hl=en
- Supported source types, source size, source selection, and import behavior: https://support.google.com/gemininotebook/answer/16215270?hl=en
- Account tiers and current usage limits: https://support.google.com/gemininotebook/answer/16213268?hl=en
- Presenter and detailed Slide Deck generation: https://support.google.com/gemininotebook/answer/16757456?hl=en