ZEVQORA — "That's why we built ZEVQORA" launch film

A self-contained, zero-budget production project for the ZEVQORA launch film. It is built with Remotion and local audio tools, and rendered entirely on the local machine.

This repository is separate from the ZEVQORA product (ZEVQORA/Zevqora) and the website design repo (ZEVQORA/Web_Design). Nothing here imports from them or deploys anything. Brand geometry and fonts were copied in from Web_Design unchanged.

Not for publication without founder approval. The rendered films are internal drafts for review.

Deliverables

File	What
out/ZEVQORA-launch-film-1080p.mp4	16:9 master — 1920×1080, 30 fps, H.264 High, AAC stereo 48 kHz, 64.0 s
out/ZEVQORA-teaser-9x16.mp4	9:16 social cutdown — 1080×1920, 30 fps, 19.0 s
public/audio/mix/{film,teaser}-mix.wav	Final stereo mixes (48 kHz / 24-bit, −14 LUFS, ≤ −1 dBTP)
public/audio/stems/{film,teaser}-{vo,music,sfx}.wav	Post-fader stems (vo + music + sfx = pre-limiter mix)
public/audio/vo/v01…v22.wav + manifest.json	Individual narration lines
public/audio/music/score-*.wav, public/audio/sfx/sfx-*.wav	Raw score and SFX layers
out/stills/	QA stills and contact sheets
docs/STORYBOARD.md	Scene-by-scene storyboard and timeline
docs/AUDIO-LICENSES.md	Licence manifest for voice, music, SFX, fonts and brand assets
docs/RENDER-QA.md	Render and QA report

Project layout

src/
  Root.tsx                    compositions: LaunchFilm (16:9), VerticalTeaser (9:16)
  compositions/               LaunchFilm.tsx, VerticalTeaser.tsx
  scenes/                     01-Opening … 07-BrandFinale (every scene takes absolute frame t)
  components/                 KineticText, ExecutionGraph, DecisionCard, CameraRig,
                              AnimatedLogo, Backdrop, Chrome (chips/panels/labels), FontGate
  config/
    timeline.json             ★ single source of truth for ALL timing (seconds)
    timeline.ts               frame helpers: ev(), voStart(), wordFrames()
    words.json                measured per-word timings of the narration
    brand.ts                  palette, fonts, easing (from Web_Design globals.css)
    graphs.ts                 deterministic graph data
  brand/marks.ts              locked logo geometry (verbatim from Web_Design)
scripts/
  narration.json              narration text, voice, speed, locked brand phonemes
  generate-voice.py           Kokoro-82M (kokoro-onnx) narration generator
  asr-check.py                local Whisper intelligibility QA
  word-timings.py             word-level timings -> src/config/words.json
  compose-music.py, dsp.py    original score synthesis
  sound-design.py             SFX layer, placed on timeline events
  mix-audio.py                ducking, VO compression, limiter, loudness, stems
  render.mjs                  stills / previews / final renders
  contact-sheet.py, qa-check.sh

Rebuild from scratch (all free, all local)

Requirements: Node 18+ (22 used), Python 3.10+ (3.13 used), FFmpeg. A Chrome headless shell is found automatically at the path in remotion.config.ts, or set REMOTION_BROWSER.

npm ci

python3 -m venv .tts/venv && .tts/venv/bin/pip install kokoro-onnx==0.6.1 soundfile numpy scipy faster-whisper==1.1.1 pyloudnorm requests pillow

mkdir -p .tts/models && cd .tts/models && curl -LO https://github.com/thewh1teagle/kokoro-onnx/releases/download/model-files-v1.0/kokoro-v1.0.onnx && curl -LO https://github.com/thewh1teagle/kokoro-onnx/releases/download/model-files-v1.0/voices-v1.0.bin && cd ../..

Then run the pipeline in order:

.tts/venv/bin/python scripts/generate-voice.py final

.tts/venv/bin/python scripts/word-timings.py && .tts/venv/bin/python scripts/asr-check.py final

for w in film teaser; do .tts/venv/bin/python scripts/compose-music.py $w && .tts/venv/bin/python scripts/sound-design.py $w && .tts/venv/bin/python scripts/mix-audio.py $w; done

node scripts/render.mjs all

bash scripts/qa-check.sh out/ZEVQORA-launch-film-1080p.mp4

Common edits

• Change a narration line: edit scripts/narration.json, then re-run the whole pipeline above. Check the new line length against the next start time in src/config/timeline.json (the VO durations are in public/audio/vo/manifest.json).
• Move a beat: change its seconds in src/config/timeline.json. Picture, score, SFX and mix all follow; re-run music, SFX, mix and render.
• Try another voice: set "voice" in scripts/narration.json (e.g. bf_emma, am_michael). Run generate-voice.py compare to audition six voices.
• Live preview / scrubbing: npm run studio.
• Fast draft render: node scripts/render.mjs preview film (half resolution).
• Stills for review: node scripts/render.mjs stills film 300,900,1500.

Brand and honesty guardrails baked in

• The Z mark is always Soft Blue #468BFF; the wordmark is Cloud on ink. The geometry is never redrawn.
• The palette is limited to Ink / Cloud / Soft Blue / Stone, plus the site's desaturated state colours.
• Every product-like screen is labelled "Illustrative visualization · not customer data".
• The film contains no metrics, customers or savings figures. VERIFIED means "ready for human review", never deployed.