اعمل فيديوهات طويلة وترويجية باستخدام Claude وHyperFrames

You are a senior Arabic motion designer, creative director, and HyperFrames engineer. Build and render a complete premium Arabic brand film from the supplied asset pack.

There is deliberately **no reference video**. Do not ask for one. The asset pack and specification below are the source of truth and are designed to produce a highly consistent result.

Do not stop at a plan, storyboard, or code sample. Create the actual HyperFrames project, transcribe and align the narration, animate the scenes, validate real rendered frames, and export the final MP4.

## Supplied pack

The workspace contains:

- `audio/VOICEOVER.wav` — final narration and master clock. If the filename differs, use the only WAV inside `audio/`.
- `fonts/` — one or more supplied font files.
- `brand/naqla-mark-blue.svg` — primary reusable blue symbol.
- `brand/naqla-logo-blue.svg` — default fictional horizontal lockup.
- `brand/brand-config.json` — brand source of truth.
- `brand/brand-tokens.css` — exact palette.

Do not request a logo, SRT, transcript, images, brand colors, or a reference video.

## Brand policy

Read `brand/brand-config.json` before designing.

Default fictional brand:

- Arabic: `نَقْلة`
- Latin: `NAQLA`
- Tagline: `من الفكرة إلى التطبيق`
- Safe fictional URL: `naqla.example`

If the voiceover clearly names another brand, replace only the written Arabic/Latin brand name with the spoken name. Continue using the supplied blue symbol, exact color palette, layout system, character, and motion direction so the output remains visually consistent.

If the voiceover does not clearly state a brand name, use `نَقْلة`. Never guess a real domain. Use `naqla.example` only with the default fictional brand; otherwise show no URL unless one is spoken.

Use `naqla-mark-blue.svg` as the authoritative symbol. For a detected brand name, build a live-text lockup using the same symbol plus the selected font. Do not redraw, recolor, distort, or place the mark over the mascot.

## Source-of-truth rules

- The WAV is the master clock and the only content authority.
- Never rewrite, regenerate, trim, stretch, denoise, reorder, or replace it.
- Inspect duration, waveform, sample rate, channels, loudness, head/tail silence, and speech regions.
- Transcribe the Arabic accurately, preserving any spoken English product terms.
- Create local `transcript.srt` and `transcript.json` files with cue timing.
- Correct obvious Arabic recognition errors by checking the waveform/context.
- Use only spoken or unambiguously implied claims.
- Visible text must be short editorial headlines, not full burned-in subtitles.
- If a word is uncertain, do not enlarge it as a headline.

## HyperFrames contract

Read and follow all available HyperFrames skills relevant to project creation, core composition, animation, creative direction, keyframes, Studio structure, audio/media, CLI validation, and rendering.

Technical target:

- 1280×720, landscape 16:9.
- 30 fps.
- Duration exactly matches the WAV at the correct final-frame boundary.
- H.264 video with AAC audio.
- Arabic RTL.
- Final filename: `FINAL_NAQLA_ARABIC_BRAND_FILM.mp4`.
- All media local and frozen.
- One paused, seek-safe master timeline driven only by HyperFrames composition time.

Forbidden:

- visual animation using `setTimeout`, `setInterval`, or free-running `requestAnimationFrame`;
- `Date.now()`, `performance.now()`, or unseeded randomness;
- standalone autoplaying GSAP/CSS/WAAPI timelines;
- runtime network requests, remote fonts, stock imagery, or external logos;
- AI-generated raster characters that change identity;
- fake browser recordings or fake real-product screenshots;
- repetitive PowerPoint fades or full-screen subtitle slides;
- hiding real collisions with blanket overlap allowances.

## Font selection

Inspect all files inside `fonts/` before choosing them.

- Read family, weight, style, and Unicode coverage.
- Confirm Arabic glyph coverage and correct browser shaping.
- Prefer a supplied family with regular and bold/display weights.
- Use Latin-only fonts only for Latin HUD/timecode, never Arabic.
- If no supplied font supports Arabic, use local `Tahoma`, then `Arial`, and record the fallback in `QA_REPORT.md`.
- Copy chosen files into project assets and load them through local `@font-face`.
- Never synthesize fake bold when a real weight exists.

Render and inspect this font proof before scene production:

`نَقْلة — من الفكرة إلى التطبيق — أدوات — تطبيق — أسرع — ١٢٣٤٥`

Reject any font showing separated letters, reversed RTL order, missing dots/diacritics, tofu, or clipped glyphs.

## Fixed visual identity

Use the exact palette from `brand/brand-tokens.css`:

- paper `#F5F1E8`
- surface `#FBFAF6`
- ink `#11191F`
- deep blue `#0B3155`
- primary blue `#1769C2`
- cyan `#21A7C4`
- orange `#F26A0A`
- yellow `#FFB000`
- mint `#CDE9DF`
- warm gray `#898781`
- hairline `#D8D2C7`

Approximate full-film color balance:

- 65–75% cream/surface;
- 15–20% ink;
- 6–10% blue/deep blue;
- 3–6% orange;
- cyan/yellow/mint only as support accents.

The blue logo is the primary brand anchor. Orange is an action/accent color, not a competing logo color.

Creative feel:

- premium Arabic editorial-tech campaign;
- warm printed paper;
- bold Arabic typography;
- black one-bit halftone illustration;
- modular interface cards;
- thin rules and registration marks;
- fast semantic edits;
- playful but controlled character acting;
- generous negative space.

Avoid corporate stock style, childish cartoons, cyberpunk/neon, glassmorphism, glossy 3D, random gradients, excessive shadows, and generic SaaS templates.

Backgrounds:

- Default: stable paper-cream field.
- Add a very subtle deterministic dot grid/print texture.
- No random texture flicker.
- Use a near-black background for only one brief contrast scene near the end.

## Typography and HUD

- Arabic headlines: bold, compact, usually two to four lines.
- Highlight one key word in blue, cyan, or orange.
- UI text: clear medium weight.
- Preserve Arabic shaping, ligatures, dots, punctuation, and RTL order.
- Never stretch text non-uniformly.
- Keep headline lines generally below 650px width.
- Maintain at least 18px painted-pixel clearance between animated lines.
- Large emphasized words require their own explicit block, line height, and margin.
- Normal entrance overscale: 1.08–1.25; absolute maximum 1.30.
- No word may travel through another line.

Editorial HUD across most scenes:

- upper-left: `NAQLA` or detected brand plus running timecode;
- upper-right: scene number and short Arabic chapter label, e.g. `03 // الفكرة`;
- small warm-gray type;
- all pixels at least 32–48px inside canvas;
- may disappear on the contrast scene/end card.

## Original mascot

Create one reusable programmatic mascot; never generate separate images per scene.

Appearance:

- friendly young Arab guide/learner;
- rounded slightly oversized head;
- dense rounded black curly hair;
- thick simple eyebrows;
- large vertical oval black eyes with tiny cream highlights;
- small rounded nose and expressive mouth;
- small ears, optional subtle beard/stubble;
- simple shirt/shoulders;
- black-and-cream one-bit halftone/dither rendering;
- no photorealism, skin shading, detailed fingers, or glossy 3D.

Implementation:

- one shared Canvas or SVG renderer;
- stable pose controls for head/body, eyes, blink, brow, mouth, arms, point, and wave;
- identical silhouette/facial proportions throughout;
- prefer a 2px halftone step when small; use 3px only when eyes/mouth remain clear.

Performance:

- deterministic irregular blinks;
- gentle 1–3px equivalent idle movement;
- eyes anticipate arriving graphics;
- occasional nod, point, thinking pose, or wave;
- no constant bounce.

Talking mouth:

- use generated SRT cues as global speech ranges;
- while speech is active, animate `mouthOpen` with a deterministic 2.6–3.2 cycle-per-second syllable pattern;
- visible range approximately 0.08–0.70;
- 40–80ms cue-boundary envelope;
- closed/resting during silence;
- readable talk motion, not fake phoneme-perfect lip sync;
- pure function of global composition time;
- clone pose state before overriding the mouth.

## Layout safety

- absolute outer margin: 32px;
- preferred content margin: 56–72px;
- important content within x=64…1216 and y=70…660;
- 24px minimum between independent text blocks;
- 32px minimum between mascot painted pixels and text/UI;
- 48px minimum between brand mark/wordmark and mascot on final card;
- 18px minimum between large animated word and preceding line throughout its path;
- inspect painted pixels, not just canvas/DOM boxes;
- check entrance, overshoot, idle, and exit, not only final state.

## Motion language

Use two to four coordinated ideas per scene:

1. Editorial snap: 18–50px purposeful travel plus restrained settle.
2. Semantic type build: words/lines in meaning order with 40–90ms stagger.
3. Card assembly: surface/border, then icon, title, status.
4. Shape handoff: a blue/orange line, dot, underline, or card edge bridges scenes.
5. Character reaction: look/point/nod after the relevant visual lands.
6. Masked print reveal: only for one major headline/section change.

Transitions normally last 4–10 frames. Use strong ease-out curves and small settles. Do not repeat one fade-up for everything. Text must remain readable after landing. No infinite loops; idle motion must derive from composition time.

## Story extraction and scene count

Extract from the transcript:

- hook/question;
- problem or list of needs;
- brand/platform promise;
- main value;
- process/steps;
- features/topics;
- audiences/use cases;
- simplicity/time/result claim;
- recap;
- final CTA.

Build 9–13 scenes based on narration duration.

- normal scene: 2.5–4.5s;
- rapid list: 1.5–3s;
- dense explanation: up to 5.5s;
- end card: 1.8–2.8s.

Align cuts with cue/phrase boundaries. Never cut inside an important word. Merge neighboring stages for short narration; split dense middle stages for long narration. Do not fill time with meaningless idle motion.

## Required scene grammar

Follow this order so every result shares the same strong visual base. Adjust percentages to actual phrasing.

### 01 — Hook, first 0–9%

- Cream field.
- Small label such as `// سؤال سريع` or category equivalent.
- Strong two-line headline on left.
- One key word in primary blue/cyan.
- Large mascot on right facing text, with clear talking mouth.
- Label, headline lines, then mascot rise/head tilt.

### 02 — Rapid needs/options, 9–18%

- Three to five short spoken nouns/questions.
- Modular 2×2 grid or editorial stagger.
- Simple locally drawn icons.
- Cream cards with blue/orange/yellow accents.
- Each item lands with its cue using related but non-identical motion.

### 03 — Brand reveal, 18–25%

- Supplied blue mark plus live-text brand name.
- Strong spoken promise/tagline.
- Thin blue frame/registration marks.
- Logo enters around scale 0.88→1, not from zero.
- Mascot presents from opposite side with no overlap.

### 04 — Value proposition, 25–38%

- Bold statement on one side.
- Two or three conceptual interface/content cards on the other.
- Blue/cyan for trust/understanding; orange for action.
- A clear line/arrow connects problem to practical result.
- Cards must be concepts, not fake screenshots.

### 05 — Choice/process, 38–49%

- Extract three or four spoken action verbs.
- Large stacked action words plus feature chips.
- Tall rounded phone/app panel on opposite side when appropriate.
- Cards populate in spoken order.
- Small mascot sits low between zones, about 180–220px wide, always at least 32px away from copy/chips/phone.

### 06 — Capabilities/topic universe, 49–64%

- Four to six labeled nodes around central blue brand node.
- Original locally drawn SVG icons only.
- Connect nodes in spoken order with thin ink lines and blue/orange endpoints.
- Subtle deterministic reframe/map expansion; no chaotic orbit.

### 07 — Audience or before/after, 64–73%

- Two related columns or one continuous journey path.
- Use audience levels if spoken; otherwise use two use cases or before/after.
- One continuous blue/orange line connects both.
- Mascot travels/points without crossing text.

### 08 — Simplicity/result, 73–84%

- Tangled warm-gray route resolves into one direct blue/orange path.
- Theory/document blocks collapse into one practical result card.
- Strong short phrase from narration, not invented marketing copy.

### 09 — Faster outcome/browser scene, 84–91%

- Left 35–38%: three short lines ending in one oversized result word.
- Right 62–65%: conceptual browser/application panel with title, progress, milestones, tasks, and a simple halftone clock/progress graphic.
- Oversized word approximately 110–126px, own block, line-height .90–1.0, 22–28px top spacing.
- Entrance scale only 1.20–1.30 from slightly below.
- Maintain 18px painted-pixel clearance from line above at every frame.

### 10 — All-in-one recap, 91–96%

- Three core offering modules from the narration.
- Bring them from three directions into one unified rounded container.
- Blue frame, orange completion point, mint success detail.
- Mascot may help place the last module.

### 11 — Contrast statement and end card, final 4–9%

- One brief near-black frame with the user's final action phrase in cream, blue, and orange.
- Return to cream end card.
- Top hero group: blue mark/wordmark and mascot side-by-side, never overlapping.
- Start around: logo/lockup 180–230px high left-of-center; mascot 280–320px wide right-of-center.
- At least 48px between painted pixels during entrances and hold.
- Final promise below, then spoken CTA.
- URL only if spoken or default `naqla.example` is active.
- Logo enters scale .82→1; mascot rises in its own column and gives a small wave.
- Hold the final clean frame until the exact end of WAV.

## Shared visual components

Build reusable components for:

- HUD/timecode;
- Arabic headline block;
- accent word/underline;
- rounded interface card;
- topic node;
- progress bar/milestone row;
- check/status chip;
- phone shell;
- conceptual browser shell;
- simple geometric icon system;
- dotted paper field;
- brand mark/live-text lockup;
- mascot renderer.

Recommended UI constants:

- radius 20–32px;
- border 1.5–2.5px;
- shallow paper shadow only;
- padding 22–34px;
- icons 2–3px stroke;
- no glass blur.

## Project structure

Create:

- `index.html` — master composition and voice track;
- `compositions/s01-*.html` through final scene — one sub-composition per scene;
- `assets/css/tokens.css` and `shared.css`;
- `assets/js/runtime.js`, `mascot.js`, and transcript data/helper;
- `assets/fonts/`, `assets/audio/`, and `assets/brand/` with frozen copies;
- `transcript.srt` and `transcript.json`;
- `BRIEF.md`, `DESIGN.md`, and `QA_REPORT.md`.

Use one scene per sub-composition for readable Studio timelines. Use explicit in/out timing and one master voiceover track. Add no music or SFX because none are supplied.

## Workflow

1. Inventory pack files.
2. Inspect WAV and fonts.
3. Transcribe/align Arabic audio.
4. Detect brand/content facts; apply default only when needed.
5. Render/approve font proof through visual inspection.
6. Write `BRIEF.md` and `DESIGN.md` from this specification.
7. Initialize/resume HyperFrames using installed workflow.
8. Start a reversible history checkpoint labeled `naqla public brand film`.
9. Build tokens, components, mark lockup, transcript gate, and mascot first.
10. Build cue-aligned scene plan and implement scenes.
11. Inspect timeline for gaps/overlaps.
12. Run current lint/check/snapshot/transition tools.
13. Fix all real problems before render.
14. Render and verify delivery.
15. End history checkpoint after verified output.

Do not stop for aesthetic approval. Ask only if no readable WAV exists.

## Mandatory visual QA

Inspect actual rendered frames, not only code:

- first/final frame;
- every transition boundary;
- at least three internal frames per scene;
- all oversized-word paths;
- all mascot paths;
- phone/browser at densest state;
- end-card entrance and hold;
- Arabic shaping at final size.

Create:

- full-film contact sheet at about 1 fps;
- transition sheet;
- mouth strip with tightly spaced spoken frames;
- silent-gap mouth proof;
- strip for every oversized Arabic word entrance;
- mascot/copy collision strip for process/phone scene;
- four end-card frames from entrance to hold.

Acceptance:

- at least three visibly different mouth states during speech;
- resting mouth during silence;
- 32px mascot/text/UI separation;
- 48px mark/mascot end-card separation;
- 18px animated text-line separation;
- correct connected RTL Arabic;
- no clipped dots, hamzas, descenders, or diacritics;
- HUD at least 32px from canvas edge;
- no missing assets, blank/loading frames, browser chrome, scrollbars, console errors, random shimmer, scene gaps, or non-deterministic seeking.

Do not silence overlap warnings; fix geometry and motion paths.

## Render and verification

Render:

`FINAL_NAQLA_ARABIC_BRAND_FILM.mp4`

Verify:

- 1280×720;
- 30 fps;
- duration matches WAV at correct frame boundary;
- H.264 + AAC;
- voice begins at 0.0;
- no unintended tail or missing final frame;
- visual events align with transcript cues.

Write `QA_REPORT.md` with:

- WAV path/metadata;
- transcript paths;
- detected/default brand;
- font inventory/selection;
- scene timing table and visible copy;
- mascot/talk implementation;
- animation/transition map;
- validation commands;
- snapshot/contact-sheet paths;
- collision and shaping checks;
- ffprobe result;
- honest limitations.

## Definition of done

Complete only when the editable HyperFrames project, transcription, font proof, consistent identity, all scenes, deterministic mascot talk, QA images, report, and verified MP4 all exist. The result must feel like one premium Arabic editorial-tech film, not unrelated slides.

At completion report the project path, MP4 path, transcript path, QA report path, detected/default brand name, chosen Arabic font, duration, validation result, and any limitation. Do not merely say `done`.