System Requirements Document for modi-government-election

1. Introduction

modi-government-election is a documentary script production workspace for a single creator building a Hinglish political documentary on the Modi government. The product exists to turn a long, information-heavy political narrative into a disciplined, lesson-by-lesson narrator script, then into the downstream assets the creator needs to finish the film: a readable script PDF, a clone-voice narration, a source list, and a static data and image pack for editing.

The audience is the creator themselves — a documentary scriptwriter who researches and writes, a narrator/voice producer who turns the finished script into audio, and a documentary editor who assembles the final cut. The product is not a news site, not a publishing platform, and not a general-purpose writing tool. It is a focused production instrument for one documentary project, and its value is clarity, structure, and factual weight.

The documentary's subject matter is fixed and must be treated as content, not as a product feature: the Modi government's arc from 2014 optimism, through the 2016 shock, the 2019 record mandate, the 2020 pandemic, farm protests, Article 370, CAA, Agnipath, identity politics and institutional debates, the 2024 political correction, and the 2025–2026 comeback story. The script's editorial stance is explicit: neither hero nor villain, just dissection of a political journey.

2. System Overview

The system is a web application with an anonymous public entry, self-service identity establishment, and a set of protected production workspaces. A single creator account owns the durable script, source, and data records; the narrator and editor personas work inside the same project record rather than in separate tenants.

Current delivery covers eight destinations: Landing, Login, Sign Up, Script Workspace, Script Export, Voice Narration, Sources, and Data & Images. The Script Workspace is where the Hinglish narrator script is generated and revised lesson by lesson. Script Export produces the narrator-only PDF in Calibri (body) font, in paragraph format, suitable for PPT-like reading. Voice Narration produces clone-voice text-to-speech with ElevenLabs-optimized pronunciation. Sources compiles the source list with multiple news headlines, video links, and multiple source links beyond Pew Research Center. Data & Images produces the follow-up PDF with static data and images for documentary editing.

Actors are the three accepted human personas — Documentary Scriptwriter, Narrator / Voice Producer, and Documentary Editor — plus external providers: the deep research connector, the PDF plugin, the data analytics plugin, and ElevenLabs. The application owns identity, durable project state, script versioning, export rendering, and the assembly of source and data packs. External providers own research retrieval, PDF rendering assistance, data analytics, and voice synthesis.

Narrow exclusions: the product does not publish the documentary, does not host video, does not provide a general news feed, does not manage a team or multi-tenant workspace, and does not include the excluded source file 076a9f45_Modi_Government_50_70_Minute_Final_Documentary_Script.pdf as content.

2a. Product Interpretation and Delivery Boundary

The creator arrives with a documentary already in mind and a script that is difficult to read. The product's job is to make that script readable, narratable, and editable. Everything the product does is in service of that chain: write the script in Hinglish, keep the narration font size identical across every lesson, box each technique separately, export a clean narrator-only PDF in Calibri (body), generate clone-voice narration tuned for ElevenLabs pronunciation, then hand over sources and a static data and image pack.

Access ownership is application-owned. The Landing page is anonymous and explains the product before any identity is established. Sign Up and Login are the anonymous entry boundaries for identity establishment and returning verification. Script Workspace, Script Export, Voice Narration, Sources, and Data & Images are protected because they hold durable, creator-owned project state that must remain bound to the correct participant and must be resumable across sessions.

The current horizon is the full accepted chain above. Future horizons — additional documentaries, multi-project libraries, collaborative review, and publishing integrations — are not part of current delivery and are not built now.

2b. Source Content Inventory

The authoritative content sources for this project are the deep research connector, the Thinking Analysis conversation, the PDF plugin, and the data analytics plugin. The excluded PDF (076a9f45_Modi_Government_50_70_Minute_Final_Documentary_Script.pdf) is explicitly excluded by the user and supplies no content.

The documentary content that must be covered by the script, in the user's own framing, is:

• 5 June 2024: India saw another election result. Television screens were changing numbers, and one number kept appearing: 240.
• 2014: BJP had 282 seats. 2019: 303. Now: 240.
• 240 did not mean the government ended. Narendra Modi returned for a third term. NDA had 293 seats.
• One important thing had changed: BJP was now below the 272 mark for a Lok Sabha majority on its own strength.
• The voter had neither removed Modi from power nor given a 2019-style blank cheque.
• Open questions the documentary must address: Did public support fall? Did the economy punish the voter? Did unemployment and prices change the narrative? Did the opposition finally create a credible challenge? Or did Indian democracy simply make a familiar correction — forcing a very powerful government to negotiate a little more?
• Editorial stance: neither hero nor villain. Dissect a political journey.
• Coverage arc: 2014 optimism → 2016 shock → 2019 record mandate → 2020 pandemic → farm protests, Article 370, CAA, Agnipath, identity politics, institutional debates → 2024 unexpected political correction → 2025 and 2026 comeback story.
• Editorial logic: if you look only at achievements, the story is incomplete. If you look only at controversies, it is also incomplete. The real story is between the two.

2c. Page Content and Component Coverage

Landing

• Information/state: anonymous public entry. Full-width black band hero (100vw, 40vh) with a single oversized Archivo Black headline breaking across two lines: "SCRIPT. SOURCE. STORY." Below the band, a 1px red rule spans the viewport, then a three-column grid of white panels, each a single sentence describing one of the product's three functions — Script, Sources, Data. No image, no gradient, no button in the hero.
• Primary action: the only interactive element is a text link "Start writing" in red, underlined, sitting in the third column, leading to Sign Up.
• Supporting actions: a secondary text link to Login for returning users.
• Domain entities: product function descriptions (Script, Sources, Data).
• Component responsibilities: hero band, red rule, three-column panel grid, "Start writing" link, Login link.
• States: loading — static, no async content; empty — not applicable; success — panels render; error — not applicable; recovery — not applicable.

Login

• Information/state: returning verification surface for the creator account. Email and password fields, submit control, link to Sign Up.
• Primary action: submit credentials to establish a session.
• Supporting actions: navigate to Sign Up; recover from invalid credentials by re-entering.
• Domain entities: creator account, session.
• Component responsibilities: credential form, validation messaging, session establishment.
• States: loading — submit disabled with progress indication; empty — fields blank; success — redirect to Script Workspace; error — invalid credentials message with fields preserved; recovery — retry without losing entered email.

Sign Up

• Information/state: self-service enrollment surface. Email, password, and confirmation fields, submit control, link to Login.
• Primary action: create the creator account and establish a session.
• Supporting actions: navigate to Login; correct validation errors inline.
• Domain entities: creator account, session.
• Component responsibilities: enrollment form, validation messaging, account creation, session establishment.
• States: loading — submit disabled with progress indication; empty — fields blank; success — redirect to Script Workspace with an empty project state; error — duplicate email or weak password message; recovery — retry with fields preserved.

Script Workspace

• Information/state: the primary production surface. Fixed left sidebar (240px on desktop, collapsing to a top bar on mobile) with numbered lesson entries (01, 02, 03) in Archivo Black; the active lesson gets a red left border and its number turns red, with no animation. The script narrator text sits in a single 780px measure at 18px/1.6 Source Sans 3, with lesson headers in Archivo Black 42px uppercase above a full-width 1px black rule. Narration font size is identical across every lesson, enforced by a single CSS class. Technique boxes are rectangles with a 2px red left border and 1px black outline, each containing a single technique label in uppercase 14px with 0.08em tracking and a short Hinglish description — one box per technique, aligned to the grid in a 4-column inset. Bridges are written in Hinglish and kept to less than 1 to 2 lines. The script contains no SOURCES markers such as [S05][S06][S07] and must be understandable without them.
• Primary action: generate and revise the Hinglish narrator script lesson by lesson, with every technique captured in its own box and no tiny script strategy detail missed.
• Supporting actions: add, reorder, and select lessons; edit narrator paragraphs; add and edit technique boxes; add and edit short Hinglish bridges; save revisions.
• Domain entities: project, lesson, narrator paragraph, technique box, bridge, script revision.
• Component responsibilities: lesson sidebar with numbered entries and active marker, lesson header with rule, narration measure, technique box grid, bridge field, save and revision controls.
• States: loading — lesson list and script content loading; empty — new project with no lessons, prompting the first lesson; success — script saved and lesson state updated; error — save failure with retry and unsaved-change preservation; recovery — restore last saved revision.

Script Export

• Information/state: export surface for the narrator-only script. Shows the export configuration and a preview of the paragraph-format, PPT-like layout. The exported PDF contains only narrator words, organized lesson by lesson, in Calibri (body) font, with identical narration font size throughout.
• Primary action: generate and download the narrator-only script PDF.
• Supporting actions: preview the export; re-export after revisions; return to Script Workspace.
• Domain entities: export job, script PDF, lesson ordering.
• Component responsibilities: export configuration, preview renderer, PDF generation, download control.
• States: loading — export job in progress; empty — no script content to export, prompting a return to Script Workspace; success — PDF available for download; error — export failure with retry; recovery — regenerate from the latest saved revision.

Voice Narration

• Information/state: narration surface. Shows the script paragraphs queued for narration, the connected clone voice, and the ElevenLabs pronunciation tuning state.
• Primary action: generate clone-voice text-to-speech of the script with ElevenLabs-optimized pronunciation.
• Supporting actions: connect or select the clone voice; adjust pronunciation tuning; preview a paragraph; regenerate a paragraph; download the narration audio.
• Domain entities: clone voice, narration job, narration audio, pronunciation tuning profile.
• Component responsibilities: voice connection control, paragraph queue, pronunciation tuning panel, generation control, audio preview and download.
• States: loading — narration job in progress per paragraph; empty — no clone voice connected or no script paragraphs, prompting the required setup; success — audio generated and playable; error — voice connection failure or generation failure with retry; recovery — regenerate individual paragraphs without losing the rest.

Sources

• Information/state: source compilation surface. Tabular alignment with hairline black rules between rows: label left in muted grey, value right in black. Rows carry multiple news headlines, video links, and multiple source links beyond Pew Research Center.
• Primary action: compile and deliver the source list for the documentary.
• Supporting actions: add a source manually; retrieve sources through the deep research connector; edit or remove a source; reorder sources; export the source list.
• Domain entities: source entry, headline, video link, source link, source category.
• Component responsibilities: source table, add and edit controls, research retrieval control, export control.
• States: loading — retrieval in progress; empty — no sources yet, prompting retrieval or manual entry; success — source list saved and exportable; error — retrieval failure with retry; recovery — manual entry remains available when retrieval fails.

Data & Images

• Information/state: the follow-up pack surface. Tabular alignment with hairline black rules between rows. Contains static data and images for documentary editing, with source data represented as simple bar charts and tables drawn with CSS — black bars on off-white, red for the single highlighted data point. Documentary thumbnails use high-contrast black-and-white crops with a red registration mark overlay.
• Primary action: produce the follow-up PDF containing static data and images for documentary editing.
• Supporting actions: add or import static data points; attach images; preview charts and tables; reorder items; export the pack.
• Domain entities: data point, chart, table, image asset, pack PDF.
• Component responsibilities: data table, chart renderer, image attachment list, preview, export control.
• States: loading — data and image processing in progress; empty — no data or images yet, prompting import; success — pack PDF available for download; error — import or export failure with retry; recovery — partial pack export with the failed items flagged.

3. Functional Requirements

FR-01 — Documentary script coverage (explicit) As a Documentary Scriptwriter, I should produce a documentary script on the Modi government covering 2014 optimism, the 2016 shock, the 2019 record mandate, the 2020 pandemic, farm protests, Article 370, CAA, Agnipath, identity politics, institutional debates, the 2024 political correction, and the 2025–2026 comeback story, so that the documentary's full arc is written.

• Trigger/input: the writer opens Script Workspace and creates or selects a lesson.
• Observable result: each listed period and theme has script content in the project.
• Access state: login required.
• Failure/recovery: unsaved lesson content is preserved on save failure and can be retried.
• Continuation: the writer moves to the next lesson.
• Owner: Script Workspace.

FR-02 — Hinglish script language (explicit) As a Documentary Scriptwriter, I should write the script always in Hinglish without hard English words and without multi-line English sentences unless required, so that the narration sounds natural to the audience.

• Trigger/input: the writer types or generates narrator text.
• Observable result: narrator text is in Hinglish and avoids hard English words and multi-line English sentences except where required.
• Access state: login required.
• Failure/recovery: the writer can revise any paragraph that drifts into hard English.
• Continuation: the writer continues to the next paragraph.
• Owner: Script Workspace.

FR-03 — Hinglish bridges, kept short (explicit) As a Documentary Scriptwriter, I should write bridges in Hinglish and keep them to less than 1 to 2 lines, so that transitions stay tight and consistent with the narration.

• Trigger/input: the writer adds a bridge between lessons or segments.
• Observable result: the bridge is in Hinglish and is less than 1 to 2 lines.
• Access state: login required.
• Failure/recovery: an over-long bridge is flagged and can be shortened.
• Continuation: the writer returns to the lesson body.
• Owner: Script Workspace.

FR-04 — Identical narration font size (explicit) As a Documentary Scriptwriter, I should have the narration font size be the same throughout, so that the script reads uniformly across every lesson.

• Trigger/input: the writer views or exports the script.
• Observable result: narration text renders at one identical size in every lesson, enforced by a single CSS class.
• Access state: login required.
• Failure/recovery: any lesson rendering at a different size is corrected to the shared class.
• Continuation: the writer continues reviewing lessons.
• Owner: Script Workspace.

FR-05 — One technique per box (explicit) As a Documentary Scriptwriter, I should place each technique in its own box, so that every technique is visually distinct and scannable.

• Trigger/input: the writer adds a technique.
• Observable result: the technique appears in its own rectangle with a 2px red left border and 1px black outline, containing a single uppercase 14px label with 0.08em tracking and a short Hinglish description.
• Access state: login required.
• Failure/recovery: a technique added without its own box is split into a separate box.
• Continuation: the writer adds the next technique.
• Owner: Script Workspace.

FR-06 — No SOURCES markers in the script (explicit) As a Documentary Scriptwriter, I should not include SOURCES markers like [S05][S06][S07] in the script, so that the script is understandable without them.

• Trigger/input: the writer reviews or exports the script.
• Observable result: no [Sxx] markers appear in the script body or the exported PDF.
• Access state: login required.
• Failure/recovery: any marker found is removed and the sentence is rewritten to stand alone.
• Continuation: the writer continues reviewing.
• Owner: Script Workspace and Script Export.

FR-07 — No tiny script strategy missed (explicit) As a Documentary Scriptwriter, I should not miss even a tiny script strategy detail, so that the script's technique layer is complete.

• Trigger/input: the writer reviews the script against the strategy.
• Observable result: every strategy detail is represented, either in narration or in its own technique box.
• Access state: login required.
• Failure/recovery: a missing strategy detail is added as a technique box or narration line.
• Continuation: the writer proceeds to export.
• Owner: Script Workspace.

FR-08 — Multiple headlines, video links, and source links (explicit) As a Documentary Editor, I should use multiple news headlines, video links, and multiple source links, not only Pew Research Center, so that the documentary's sourcing is broad and verifiable.

• Trigger/input: the editor opens Sources and retrieves or adds sources.
• Observable result: the source list contains multiple news headlines, multiple video links, and multiple source links from more than one outlet, with Pew Research Center as at most one of them.
• Access state: login required.
• Failure/recovery: if retrieval fails, manual entry remains available and the failure is reported with retry.
• Continuation: the editor exports the source list.
• Owner: Sources, with retrieval through the deep research connector.

FR-09 — Narrator-only script PDF by lesson (explicit) As a Documentary Scriptwriter, I should receive the script as a PDF containing only narrator words, organized lesson by lesson, so that the narrator reads only what must be spoken.

• Trigger/input: the writer triggers export from Script Export.
• Observable result: the PDF contains only narrator words, grouped and ordered by lesson, with no technique labels, no source markers, and no production notes.
• Access state: login required.
• Failure/recovery: export failure is reported with retry, and the latest saved revision is used on retry.
• Continuation: the writer downloads the PDF and proceeds to Voice Narration.
• Owner: Script Export.

FR-10 — Calibri (body) font in the PDF (explicit) As a Documentary Scriptwriter, I should have the PDF use Calibri (body) font because the current PDF is difficult to read, so that the script is comfortable to read.

• Trigger/input: the writer exports the script PDF.
• Observable result: the exported PDF renders body text in Calibri.
• Access state: login required.
• Failure/recovery: if Calibri is unavailable in the rendering environment, the export reports the substitution rather than silently changing the font.
• Continuation: the writer downloads the PDF.
• Owner: Script Export.

FR-11 — Paragraph format, PPT-like (explicit) As a Documentary Scriptwriter, I should have the script presented in paragraph format, suitable for a PPT-like PDF, so that the script reads as flowing narration rather than fragmented lines.

• Trigger/input: the writer exports or previews the script.
• Observable result: narrator content renders as paragraphs in a PPT-like layout.
• Access state: login required.
• Failure/recovery: fragmented output is re-rendered as paragraphs.
• Continuation: the writer downloads the PDF.
• Owner: Script Export.

FR-12 — Clone-voice text-to-speech (explicit) As a Narrator / Voice Producer, I should get text-to-speech of my clone voice with perfect pronunciation tuned for ElevenLabs, so that the narration sounds like the creator.

• Trigger/input: the narrator connects the clone voice and starts generation.
• Observable result: narration audio is produced in the clone voice with the pronunciation tuning applied.
• Access state: login required; a clone voice must be connected.
• Failure/recovery: voice connection or generation failure is reported with retry, and individual paragraphs can be regenerated without losing the rest.
• Continuation: the narrator previews and downloads the audio.
• Owner: Voice Narration, with synthesis by ElevenLabs.

FR-13 — Sources delivered after the script (explicit) As a Documentary Editor, I should receive the sources after the script, so that the delivery order matches the production workflow.

• Trigger/input: the script has been delivered.
• Observable result: the Sources surface holds the compiled source list, available after the script export.
• Access state: login required.
• Failure/recovery: if the script is not yet exported, the editor is prompted to complete the script first.
• Continuation: the editor moves to Data & Images.
• Owner: Sources.

FR-14 — Static data and images PDF next (explicit) As a Documentary Editor, I should receive a next PDF with static data and images that help me edit the documentary, so that the edit has visual and factual support.

• Trigger/input: the editor opens Data & Images after the sources.
• Observable result: a PDF containing static data and images is produced for editing support.
• Access state: login required.
• Failure/recovery: import or export failure is reported with retry, and a partial pack can be exported with failed items flagged.
• Continuation: the editor downloads the pack and begins the edit.
• Owner: Data & Images, with data analytics support from the data analytics plugin.

FR-15 — Identity establishment and returning verification (required_inference) As a Documentary Scriptwriter, I should sign up or log in to access the protected script, source, and data/image workspaces, so that my durable project state stays bound to me and is resumable.

• Trigger/input: the writer selects "Start writing" on Landing or opens a protected destination without a session.
• Observable result: a session is established and the protected destination becomes available.
• Access state: anonymous entry at Sign Up and Login; protected destinations require login.
• Failure/recovery: invalid credentials or duplicate email are reported with fields preserved for retry.
• Continuation: the writer lands in Script Workspace.
• Owner: Sign Up and Login.

FR-16 — Clone voice connection (required_inference) As a Narrator / Voice Producer, I should provide or connect a clone voice before narration, so that text-to-speech can be generated in the creator's voice.

• Trigger/input: the narrator opens Voice Narration without a connected voice.
• Observable result: the clone voice is connected and available for generation.
• Access state: login required.
• Failure/recovery: connection failure is reported with retry, and the script paragraphs remain queued.
• Continuation: the narrator starts generation.
• Owner: Voice Narration.

FR-17 — ElevenLabs access or configuration (required_inference) As a Narrator / Voice Producer, I should have ElevenLabs access or configuration available for pronunciation tuning, so that pronunciation can be tuned for perfect output.

• Trigger/input: the narrator opens the pronunciation tuning panel.
• Observable result: ElevenLabs configuration is present and tuning can be applied to the narration.
• Access state: login required.
• Failure/recovery: missing or invalid configuration is reported with guidance and retry.
• Continuation: the narrator generates narration with tuning applied.
• Owner: Voice Narration, with configuration owned by ElevenLabs.

4. User Personas

Documentary Scriptwriter

The scriptwriter is the originator of the project and the person who carries the documentary's editorial stance. Their context is a long political narrative that must cover 2014 optimism, the 2016 shock, the 2019 record mandate, the 2020 pandemic, farm protests, Article 370, CAA, Agnipath, identity politics, institutional debates, the 2024 political correction, and the 2025–2026 comeback story — without turning the film into either a defence or an attack. Their primary goal is a complete, readable, lesson-by-lesson Hinglish narrator script that is ready to be spoken.

Their distinct responsibilities are the ones only they can do: choosing what each lesson covers, writing narration in Hinglish without hard English words and without multi-line English sentences unless required, writing bridges in Hinglish kept to less than 1 to 2 lines, placing each technique in its own box, and making sure not even a tiny script strategy detail is missed. They also own the readability decision — the script must be understandable without SOURCES markers, and the narration font size must be identical throughout.

Their inputs are the research material gathered through the deep research connector and the Thinking Analysis conversation, and their decisions are editorial: which period opens a lesson, where a bridge belongs, which technique deserves its own box, and whether a sentence has drifted into hard English. They interact with the narrator by handing over the exported narrator-only PDF, and with the editor by handing over the source list and the data and image pack. Their observable success is a complete script PDF in Calibri (body) font, in paragraph format, containing only narrator words, organized lesson by lesson.

Narrator / Voice Producer

The narrator is the voice of the documentary and the person who turns the written script into audio. Their context is a finished narrator-only script that must be spoken in the creator's own voice, which means the narration is not a generic read but a clone-voice performance. Their primary goal is a clean narration audio aligned to the script paragraphs, with pronunciation tuned so nothing sounds wrong.

Their distinct responsibilities are voice-side, not writing-side: connecting or providing the clone voice, confirming ElevenLabs access or configuration, tuning pronunciation for perfect output, generating text-to-speech paragraph by paragraph, previewing, and regenerating individual paragraphs when a pronunciation is off. They do not rewrite the script; when a line reads badly aloud, they raise it back to the scriptwriter rather than editing narration themselves.

Their inputs are the exported narrator-only PDF and the connected clone voice. Their decisions are about pronunciation and pacing per paragraph. They interact with the scriptwriter by requesting a script fix when a line cannot be spoken cleanly, and with the editor by delivering the finished narration audio. Their observable success is narration audio in the clone voice, aligned to the script paragraphs, with pronunciation tuned for ElevenLabs.

Documentary Editor

The editor is the person who assembles the final cut and needs the supporting material to do it. Their context is a documentary that must be factually grounded and visually supported, which means they cannot work from the script alone. Their primary goal is an edit-ready package of script, sources, data, and visuals.

Their distinct responsibilities are sourcing and assembly: compiling multiple news headlines, video links, and multiple source links beyond Pew Research Center; checking that the sourcing is broad rather than resting on a single outlet; and gathering the static data and images that will appear on screen. They also decide which data point is the single highlighted one in a chart and which images are usable as high-contrast black-and-white crops.

Their inputs are the delivered script, the compiled source list, and the static data and image pack. Their decisions are about which sources carry which claim and which visuals support which segment. They interact with the scriptwriter when a claim needs a stronger source, and with the narrator when audio timing must match a visual beat. Their observable success is a downloaded source list and a data and image PDF that let them begin the edit without going back to the writer for missing material.

5. Core User Flows

Flow 1 — Documentary Scriptwriter: establish access and open the workspace

1. The scriptwriter lands on Landing and reads the black band headline "SCRIPT. SOURCE. STORY." and the three white panels describing Script, Sources, and Data.
2. The scriptwriter selects the red underlined text link "Start writing" in the third column.
3. Sign Up opens with email, password, and confirmation fields. The scriptwriter enters credentials and submits.
4. On success, the session is established and the scriptwriter is redirected to Script Workspace with an empty project state prompting the first lesson.
5. If the email is already registered or the password is weak, the error is shown inline with the entered email preserved, and the scriptwriter corrects and resubmits.
6. A returning scriptwriter instead uses the Login link, submits credentials on Login, and lands in Script Workspace with the last saved project restored.

Flow 2 — Documentary Scriptwriter: write the Hinglish narrator script lesson by lesson

1. In Script Workspace, the scriptwriter creates the first lesson. The lesson header renders in Archivo Black 42px uppercase above a full-width 1px black rule, and the lesson appears in the fixed left sidebar as numbered entry 01.
2. The scriptwriter writes the narrator paragraphs in Hinglish, avoiding hard English words and avoiding multi-line English sentences unless required. The narration renders in the single 780px measure at 18px/1.6 Source Sans 3, at the same size in every lesson.
3. The scriptwriter works through the documentary arc: 2014 optimism, the 2016 shock, the 2019 record mandate, the 2020 pandemic, farm protests, Article 370, CAA, Agnipath, identity politics, institutional debates, the 2024 political correction, and the 2025–2026 comeback story.
4. For the 2024 correction, the scriptwriter writes the narration around the 5 June 2024 result: the television screens changing numbers, the number 240, the comparison with 282 in 2014 and 303 in 2019, the fact that the government did not end and Narendra Modi returned for a third term with the NDA at 293 seats, and the key change that the BJP was now below the 272 mark on its own strength. The narration carries the open questions — did public support fall, did the economy punish the voter, did unemployment and prices change the narrative, did the opposition finally create a credible challenge, or did Indian democracy make a familiar correction — and states the editorial stance that the documentary will make Modi government neither hero nor villain.
5. Wherever a technique appears, the scriptwriter adds it as its own box: a rectangle with a 2px red left border and 1px black outline, a single uppercase 14px label with 0.08em tracking, and a short Hinglish description. No technique shares a box with another.
6. Between lessons and segments, the scriptwriter adds bridges in Hinglish, each kept to less than 1 to 2 lines.

7. The scriptwriter reviews the script to confirm no SOURCES markers such as [S05][S06][S07] remain and that the script is understandable without them, and that no tiny script strategy detail has been missed.
8. The scriptwriter saves. On save failure, the unsaved changes are preserved and the save can be retried; the last saved revision can be restored.
9. The scriptwriter selects the next lesson in the sidebar. The active lesson marker changes instantly — a red left border and a red number — with no animation.

Flow 3 — Documentary Scriptwriter: export the narrator-only script PDF

1. From Script Workspace, the scriptwriter opens Script Export.
2. The export surface shows the configuration and a preview of the paragraph-format, PPT-like layout.
3. The scriptwriter triggers the export. The PDF is generated containing only narrator words, organized lesson by lesson, in Calibri (body) font, with the narration font size identical throughout.
4. If there is no script content, the surface shows an empty state prompting a return to Script Workspace.
5. If export fails, the failure is reported with retry, and the retry uses the latest saved revision.
6. On success, the scriptwriter downloads the PDF and proceeds to narration.

Flow 4 — Narrator / Voice Producer: connect the clone voice and tune pronunciation

1. The narrator opens Voice Narration and sees the script paragraphs queued for narration.
2. No clone voice is connected, so the surface shows the required setup state. The narrator provides or connects the clone voice.
3. The narrator confirms ElevenLabs access or configuration in the pronunciation tuning panel.
4. If the voice connection fails or the configuration is invalid, the failure is reported with guidance and retry, and the paragraph queue is preserved.
5. On success, the clone voice and tuning profile are available for generation.

Flow 5 — Narrator / Voice Producer: generate and deliver the narration

1. With the clone voice connected, the narrator starts generation in Voice Narration.
2. Narration is produced paragraph by paragraph in the clone voice with the pronunciation tuning applied.
3. The narrator previews a paragraph. If a pronunciation is wrong, the narrator adjusts the tuning and regenerates that paragraph only; the other paragraphs are unaffected.
4. If generation fails for a paragraph, the failure is reported with retry for that paragraph.
5. On success, the narrator downloads the narration audio aligned to the script paragraphs and hands it to the editor.

Flow 6 — Documentary Editor: compile the sources after the script

1. After the script has been delivered, the editor opens Sources.
2. If the script has not yet been exported, the editor is prompted to complete the script first.
3. The editor retrieves sources through the deep research connector and adds sources manually where needed.
4. The source list is built with multiple news headlines, video links, and multiple source links, not only Pew Research Center. Rows render in tabular alignment with hairline black rules, label left in muted grey and value right in black.
5. If retrieval fails, the failure is reported with retry, and manual entry remains available so the editor can continue.
6. The editor edits, removes, and reorders sources, then exports the source list.

Flow 7 — Documentary Editor: produce the static data and images PDF

1. After the sources, the editor opens Data & Images.
2. The editor adds or imports static data points and attaches images for documentary editing.
3. Source data renders as simple bar charts and tables drawn with CSS — black bars on off-white, with red for the single highlighted data point. Documentary thumbnails render as high-contrast black-and-white crops with a red registration mark overlay.
4. The editor previews the charts, tables, and images and reorders items as needed.
5. If an import or export fails, the failure is reported with retry, and a partial pack can be exported with the failed items flagged.
6. On success, the editor downloads the data and image PDF and begins the edit.

6. Visuals Colors and Theme

The visual system follows the supplied creative direction: Swiss grid rigour after Josef Müller-Brockmann, built for information-heavy, serious content. The register is serious, analytical, and civic-minded — not entertainment, not activism.

Palette (light mode)

No gradients, no shadows, no colour beyond this system. Exactly one accent colour.

Typography

• Headings: Archivo — Archivo Black for lesson titles and section headers, uppercase, tight tracking (-0.02em), flush-left ragged-right, weight 900 at display sizes and weight 600 for subheads. No italics, no decorative weights.
• Body: Source Sans 3.
• Scale: 1.333 modular — 56/42/32/24/18/16/14 desktop, 36/28/22/18/16/14/12 mobile.
• Display headlines: clamp(36px, 5vw, 56px).
• Body narration: fixed at 18px desktop / 16px mobile, identical across all lessons per the user requirement.
• Technique box labels: 14px uppercase with 0.08em tracking.

Shape language

Hard edges only. Zero border-radius. 1px black rules as section dividers. Technique boxes are rectangles with a 2px red left border and 1px black outline. No rounded corners, no soft shadows, no organic shapes. Every element sits on the grid.

Layout

12-column modular grid with 24px gutters at desktop and 16px at mobile. Content max-width 780px for the narration measure (65–75 characters per line). Lesson headers span the full grid width with a horizontal rule beneath. Technique boxes sit in a 4-column inset, aligned to the grid. Sidebar navigation is 240px fixed on desktop and collapses to a top bar on mobile. Sources and data sections use tabular alignment: label left, value right, hairline rules between rows.

Imagery

Diagrammatic only. No photography in the workspace. Source data is represented as simple bar charts and tables drawn with CSS — black bars on off-white, red for the single highlighted data point. Technique boxes may include a small line-art icon (16px stroke, black) representing the technique type. Documentary thumbnails in the export view use high-contrast black-and-white crops with a red registration mark overlay.

Readability at every viewport

Headlines, wordmarks, labels, numbers, and card text and controls stay entirely inside the viewport and their container at 375px, 768px, and 1280px, wrapping or scaling with clamp(...) and its mobile size. No element covers any part of readable text or a control. Imagery and decoration may be cropped or bled as the direction asks, but never over readable text or controls.

7. Signature Design Concept

The public entry is a full-width black band, 100vw wide and 40vh tall, bleeding to all viewport edges. Inside it sits a single oversized headline in Archivo Black at clamp(36px, 8vw, 96px), flush-left, breaking across two lines: SCRIPT. SOURCE. STORY. There is no image, no gradient, and no button in the band — type is the architecture.

Below the black band, a 1px red rule spans the viewport. Under the rule, a three-column grid of white panels sits on the warm off-white ground with generous margins. Each panel carries a single sentence describing one of the product's three functions: Script, Sources, Data. The only interactive element in the hero is a text link, "Start writing", in red and underlined, sitting in the third column. A secondary text link to Login serves returning users.

The concept recomposes only accepted content and controls: the product's three functions, the anonymous entry, and the identity entry points. It introduces no new behaviour, page, or destination.

8. Interaction Model & Motion Direction

Interaction Model: Static Motion Tempo: restrained Hero Dimensionality: flat

Landing Hero Motion Brief

• Focal subject: the oversized Archivo Black headline "SCRIPT. SOURCE. STORY." set in the full-width black band, with the three white function panels below the red rule.
• Input → transformation → outcome thesis: the visitor arrives with no session and no context; the black band states the product's three functions as type; the red rule and the three panels resolve that statement into Script, Sources, and Data; the visitor's outcome is a single clear next action — the red "Start writing" link.
• Motion vocabulary: minimal and functional. Lesson transitions are instant state changes — no fades, no slides. Technique boxes reveal on scroll with a 120ms opacity shift. The active lesson marker in the sidebar gets a red left border that appears without animation. No parallax, no hover lifts, no decorative motion.
• Composed first frame: the black band fills the top 40vh edge to edge, the headline flush-left and broken across two lines, the 1px red rule spanning the viewport beneath it, and the three white panels sitting on the off-white ground with generous margins, the red underlined "Start writing" link in the third column.
• Reduced-motion state: all transitions are removed. The technique box reveal becomes an immediate appearance, the active lesson marker appears without animation, and the hero renders as a fully static composition with every element in its final position.

9. Non-Functional Requirements

NFR-01 — Script language constraint (explicit) The script must always be in Hinglish. Hard English words must not be used, and multi-line English sentences must not be used unless required. Rationale: the narration must sound natural to the audience.

NFR-02 — Bridge length constraint (explicit) Bridges must be in Hinglish and kept to less than 1 to 2 lines. Rationale: transitions must stay tight.

NFR-03 — Narration font size consistency (explicit) The narration font size must be identical throughout the script, enforced by a single CSS class. Rationale: uniform readability across every lesson.

NFR-04 — Technique box containment (explicit) Each technique must be contained in one box. Rationale: techniques must be visually distinct and scannable.

NFR-05 — No source markers in the script (explicit) SOURCES markers such as [S05][S06][S07] must be removed from the script, and the script must be understandable without them. Rationale: the script is read aloud and must stand alone.

NFR-06 — Script strategy completeness (explicit) Not even a tiny script strategy detail may be missed. Rationale: the technique layer is part of the product's value.

NFR-07 — Source breadth (explicit) The source list must use multiple news headlines, video links, and multiple source links, not only Pew Research Center. Rationale: sourcing must be broad and verifiable.

NFR-08 — PDF font (explicit) The script PDF must use Calibri (body) font. Rationale: the current PDF is difficult to read.

NFR-09 — PDF content scope (explicit) The script PDF must contain only narrator words, organized by lesson. Rationale: the narrator reads only what must be spoken.

NFR-10 — Narration voice and pronunciation (explicit) Text-to-speech must use the user's clone voice with ElevenLabs pronunciation tuning. Rationale: the narration must sound like the creator and pronounce correctly.

NFR-11 — Script format (explicit) The script must be in paragraph format, suitable for a PPT-like PDF. Rationale: the script must read as flowing narration.

NFR-12 — Delivery order (explicit) Sources come after the script; the static data and images PDF comes next. Rationale: the delivery order matches the production workflow.

NFR-13 — Durable project state (required_inference) Script, source, and data records must persist across sessions and remain bound to the owning creator account. Rationale: the accepted journeys require resumable, creator-owned state.

NFR-14 — Readability at all viewports (required_inference) Readable text and controls must stay whole and fully inside the viewport and their container at 375px, 768px, and 1280px. Rationale: the workspace is used across desktop and mobile.

NFR-15 — Reduced motion (required_inference) The interface must respect prefers-reduced-motion by removing all transitions and providing a usable static arrangement. Rationale: accessibility and the direction's own reduced-motion rule.

10. Tech Stack

• Frontend: React web application. Rationale: the product is a browser-based production workspace with a fixed sidebar, a narration measure, and export previews.
• Backend: Python with FastAPI. Rationale: the product requires server-side script persistence, export rendering, source retrieval, and narration job orchestration.
• Storage: Relational database for projects, lessons, narrator paragraphs, technique boxes, bridges, script revisions, source entries, data points, and image assets; object storage for generated PDFs and narration audio. Rationale: durable creator-owned state is required.
• PDF rendering: Server-side rendering using the PDF plugin, producing Calibri (body) body text in paragraph format. Rationale: the export must be a readable, narrator-only PDF.
• Voice synthesis: ElevenLabs, using the creator's clone voice with pronunciation tuning. Rationale: explicitly required by the user.
• Research retrieval: Deep research connector for multiple news headlines, video links, and multiple source links. Rationale: explicitly required by the user.
• Data analytics: Data analytics plugin for the static data used in the editing-support PDF. Rationale: explicitly required by the user.
• Containerization: Docker and docker-compose for the frontend, backend, and database. Rationale: a single shared backend serves all current APIs and domain features; no separate service is needed for the current product.

11. Assumptions and Constraints

Assumptions

• A1 (required_inference): A single creator account owns the documentary project; the narrator and editor personas work within that same project record rather than in separate tenants.
• A2 (required_inference): The creator has, or can obtain, a clone voice and ElevenLabs access or configuration before narration can be generated.
• A3 (required_inference): The deep research connector, PDF plugin, and data analytics plugin are available as content sources for retrieval, PDF rendering, and static data respectively.
• A4 (required_inference): Calibri is available in the PDF rendering environment; if it is not, the export reports the substitution rather than silently changing the font.
• A5 (explicit): The excluded file 076a9f45_Modi_Government_50_70_Minute_Final_Documentary_Script.pdf supplies no content.

Constraints

• C1 (explicit): The script must always be in Hinglish; hard English words must not be used.
• C2 (explicit): Multi-line English sentences must not be used unless required.
• C3 (explicit): Bridges must be in Hinglish and kept to less than 1 to 2 lines.
• C4 (explicit): The narration font size must be identical throughout.
• C5 (explicit): Each technique must be contained in one box.
• C6 (explicit): SOURCES markers such as [S05][S06][S07] must be removed; the script must be understandable without them.
• C7 (explicit): Not even a tiny script strategy detail may be missed.
• C8 (explicit): Multiple news headlines, video links, and multiple source links must be used, not only Pew Research Center.
• C9 (explicit): The PDF must use Calibri (body) font.
• C10 (explicit): The script PDF must contain only narrator words, organized by lesson.
• C11 (explicit): Text-to-speech must use the user's clone voice with ElevenLabs pronunciation tuning.
• C12 (explicit): The script must be in paragraph format, suitable for a PPT-like PDF.
• C13 (explicit): Sources come after the script; the static data and images PDF comes next.
• C14 (explicit): The documentary's editorial stance is neither hero nor villain; the script dissects a political journey.
• C15 (explicit): The documentary's coverage arc runs from 2014 optimism through the 2016 shock, the 2019 record mandate, the 2020 pandemic, farm protests, Article 370, CAA, Agnipath, identity politics, institutional debates, the 2024 political correction, and the 2025–2026 comeback story.
• C16 (explicit): The documentary must not rest on achievements alone or controversies alone; the real story is between the two.

Future (not current delivery)

• F1: Additional documentary projects beyond the Modi government documentary.
• F2: Multi-project libraries and cross-project search.
• F3: Collaborative review and commenting by additional participants.
• F4: Publishing or distribution integrations for the finished documentary.

12. Glossary

• Hinglish — Hindi written in Latin script, mixed with everyday English words, avoiding hard English vocabulary and long English sentences.
• Bridge — A short Hinglish transition between lessons or segments, kept to less than 1 to 2 lines.
• Technique box — A rectangle with a 2px red left border and 1px black outline containing one technique label in uppercase 14px with 0.08em tracking and a short Hinglish description. One technique per box.
• Narration measure — The single 780px column at 18px/1.6 Source Sans 3 in which narrator text is set, at the same size in every lesson.
• Narrator words — The spoken text only; the content of the narrator-only script PDF, excluding technique labels, source markers, and production notes.
• Clone voice — The creator's own voice model used for text-to-speech narration.
• ElevenLabs pronunciation tuning — The configuration applied to the clone voice so that narration pronounces correctly.
• Static data and images pack — The follow-up PDF containing static data and images that support the documentary edit.
• Political correction (2024) — The documentary's framing of the 5 June 2024 result: the BJP at 240 seats, below the 272 mark on its own strength, with the NDA at 293 seats and Narendra Modi returning for a third term.
• Editorial stance — Neither hero nor villain; the documentary dissects a political journey rather than defending or attacking it.