docs/demo-video-script.md

VoiceLedger Demo Video Script

Target length: 60-90 seconds.

0:00-0:10 — Problem

VoiceLedger is a voice-first bookkeeping app for informal sellers and home businesses. The target user records quick notes like “Sold 12 mangoes, 20 each” or “Amit owes 100” while working.

0:10-0:25 — Judge Flow and Health

On the first screen, show Guided Judge Mode and the Judge Demo Flow panel:

1. Seed demo data -> 2. Record/type -> 3. Save -> 4. View dashboard/reports

Click Seed Demo Transactions.

Show Seller Setup and mention that business name, currency preset, low-stock threshold, and language style are saved in SQLite.

Show the first-run guide and multilingual examples so judges see English, Hinglish, Gujarati-lite, Spanish, French, and Portuguese seller notes.

Open Submission Story and show the AI pipeline strip plus the “Why small models fit” card.

Click Check Demo Health from the launchpad, or open Demo Health, and show:

• Modal backend status
• Deployed backend version
• NVIDIA Nemotron parser
• SQLite database
• PDF support
• Configured Modal endpoints

Say: “The Space calls Modal first for speech and NVIDIA Nemotron parsing, with local fallback for reliability.”

0:25-0:45 — Voice or Text Transaction

Use the Sections navigation to open Record Text & Voice.

Type or speak:

Sold 12 mangoes, 20 each

Click parse. Show:

• Transcript if using voice
• Human-friendly transaction review card
• Price memory badge if a known item is missing price
• Inline review fields for correcting item, quantity, amount, customer, and notes before saving
• Warning badges for low confidence, missing fields, duplicate risk, or negative stock when present
• Parse source/status
• Language/confidence chip for multilingual notes
• Command Center update
• Voice command shortcuts such as close today, show Amit, and stock mangoes
• “Saved just now” receipt with stock, customer, or amount side effects

0:45-1:05 — Bookkeeping Updates

Use the Sections navigation to open:

• Dashboard to show sales/profit/credit and the sales/expense timeline
• Field Test to show the seller checklist and anonymized “who/tried/changed” notes
• Customer Credit to show dues, a selected customer detail view, and a follow-up message
• Customer Credit to show the debt reminder queue sorted by highest balance
• Inventory to show stock, a selected item detail view, and reorder intelligence
• Ledger to show saved transactions

Say: “The model handles messy input, but the app owns the accounting state.”

1:05-1:20 — Correction and Export

In Ledger, load a transaction by id, update or delete it, and show balances refresh.

Click Download CSV.

Open Reports & PDF, run Daily Closeout, then generate the PDF, translated WhatsApp summary, and CSV export.

Suggested screenshot/GIF moments:

• Record flow: Guided Judge Mode, first-run guide, multilingual examples, review card, price memory, inline correction, parse source, warning badges, save receipt.
• Seller setup: currency/threshold/language style and Command Center.
• Field Test: seller checklist, feedback notes, and correction log.
• Submission Story: AI pipeline and small-model fit card.
• Dashboard: sales, expenses, profit, credit, timeline, top item, low-stock inventory.
• Reports/Ledger: Daily Closeout, PDF download, WhatsApp summary, CSV export.

1:20-1:30 — Close

Close with:

“VoiceLedger turns everyday seller notes into a working ledger: voice, credit, stock, reports, and exports in one Gradio Space.”