# AEVRA R1 — reference study and editorial decisions

## Evidence and limits

Official films downloaded locally for private analysis on 2026-09-09. None of their footage, graphics or audio is used in the AEVRA film. Contact sheets, source metadata, candidate boundaries and audio measurements are in `output/automotive/commercial/reference-analysis/`.

Sources:
- Tesla, **Meet the Updated Model S & X**: https://www.youtube.com/watch?v=k0oMa2jWXaM
- Tesla, **Roadster**: https://www.youtube.com/watch?v=tw4jkyfY4HE
- Aston Martin, **The New Aston Martin Vantage | Engineered for real drivers**: https://www.youtube.com/watch?v=UbktqyPB3Rw
- Porsche, **The new 911. The one and always.**: https://www.youtube.com/watch?v=OUPQGEwO6hA

Method: decode the actual films, detect candidate cuts with FFmpeg scene score >0.15, sample every candidate shot, inspect the resulting sheets, and measure audio energy and spectral attacks. This is substantially more coverage than the earlier eight-frame study. Detection is not an edit decision list: flashes can create false positives; similar compositions and dissolves can hide cuts. In particular the Roadster opening is undersegmented. Do not present the candidate-shot medians as verified editorial statistics.

Audio limitation: this environment does not provide usable auditory perception. The measurements below describe energy and timing, not a claimed listening assessment of instrumentation, narration or mix quality. A human listening pass remains necessary.

## Tesla: primary visual reference

The updated Model S & X film gives the clearest applicable structure. Around 0–5.8 s, a relatively close exterior gives way to a distant car moving across a wide landscape. Around 5.8–11.2 s, seat hardware, hands and the driver bring the viewer inside. At 11.2–15.2 s, windshield, badge, wheel and front tracking alternate. The 15.2–18.2 s wide bend restores geography. Around 18.2–25.5 s, tighter exterior and feature shots accelerate the sequence. From approximately 25.5–32.5 s, rapidly changing driving angles keep the car moving through the environment. The final wide crossing leads into a plain mark on black around 35.4 s.

What makes that work is not a transition pack. Most visible boundaries are direct cuts. The changing scale creates contrast: a wheel fills the frame, then the entire car becomes small in the landscape. The driver and passing environment supply physical context. The car remains the brightest, most readable subject without every shot being a centered catalog image. Long establishing beats and brief inserts have different jobs.

A separate two-second interval review of Roadster confirms the detector missed opening changes: front and side beauty views occupy the first roughly nine seconds, rear/acceleration imagery follows, and moving front/side views return before the mark. That additional sheet is saved as `tesla-roadster/interval-review.jpg`. Roadster is useful for its low lens height, broad landscape, clean side/front silhouette and very restrained ending. Its wide aspect ratio makes the horizontal body line feel deliberate. Our footage should preserve that clarity without imitating Tesla's mark or tracing its exact sequence.

Application: introduce the road, interrupt with tactile details, reveal the car in motion, tighten the inserts, return to an expansive driving shot, and give the AEVRA mark a clean ending. Keep the front tracking angle readable; do not cover it with a paragraph of copy.

## Aston Martin: physical cause and response

The opening links the driver, road and pedal. Approximately 11–34 s, the engineering visualization connects controls with mechanical response. The film subsequently alternates the driver with low body, wheel, rear and wide track imagery. Around 40–43 s, extremely brief passages create a burst of intensity before longer front/rear tracking shots. The ending separates a short brand statement from the model name and final signature.

Transferable lesson: an insert should explain a sensation or action. The console belongs between driving shots because it suggests the driver's environment; repeatedly cutting to unrelated beauty angles does not build a story. The light travels across surfaces rather than appearing as an arbitrary animated overlay.

Not transferable to this production: Vantage's combustion mechanics, exhaust soundtrack or transparent drivetrain imagery. AEVRA is an electric concept and should not imply a photographed or engineered powertrain it does not have.

## Porsche: scale, atmosphere and release

The opening moves from road to aerial to hands. Daylight driving is repeatedly contrasted with dark architectural beauty shots. The roughly 54–64 s sequence tightens through road, hand control, rear, tunnel and front imagery. Around 74–85 s the road and passing bodywork become short rhythmic accents. A longer rear driving shot near 87.5–95.5 s provides release before the end card.

Useful principles: keep a coherent direction of travel; let blur arise from motion; change distance without losing geography; reserve a longer shot after a dense passage. A long final driving hold feels earned after contrast. It feels inert when the entire ad has the same pace.

## Sound and edit strategy for AEVRA

The source-analysis script measures spectral flux in 10 ms hops and logs the nearest strong audio attack to each candidate cut. A nearby attack is not automatically a musical beat: road effects, voice and transients can all produce peaks. This prevents falsely claiming that every source cut lands on a downbeat.

For our own track, the late passage of Scott Buckley's **Convergence** has a strong measured periodicity close to 90 BPM (approximately 0.667 s). At 24 fps, that is 16 frames per beat, 64 per four-beat bar. Use whole bars for the larger turns and one- or two-beat inserts for acceleration. A cut does not need to hit every beat. Continuous footage can carry several beats to keep the car readable.

Music: https://www.scottbuckley.com.au/library/convergence/
License and required attribution: https://www.scottbuckley.com.au/library/licensing/
The composer describes strings, synth and a late 4/4 kick. That description is sourced, not an assertion of auditory review here.

Use a continuous late excerpt to retain the composer's harmonic progression and natural ending, with an opening gain envelope and a controlled final tail. Layer cleaned generated road/wind beneath action. Carry road ambience a little over the picture cuts so the scene does not audibly restart every second. No artificial combustion roar. No narration, per the user's latest decision.

## Quality gates

1. Vehicle: closed roof, silver paint, correct lamps, consistent wheel/body proportions. Reject obvious mutations even if the composition is attractive.
2. Motion: the car moves forward with its lane and tires; no floating, morphing roof or crossing the centerline in a retained passage.
3. Edit: every new angle earns its place; no rejected side take, invented wheel clip or lengthy parked reveal.
4. Picture: review native frames as well as contact sheets; keep meaningful detail inside the delivery crop; match exposure gently without crushing the cabin.
5. Sound: measured beat anchors, controlled peaks, stereo 48 kHz, no discontinuities at edit points. Listening approval is separate from numeric validation.
6. Delivery: full decode, exact duration/frame count, browser playback with working sound controls, download, attribution and reproducible timeline.

This study supports a better creative decision. It does not establish that the resulting generated footage matches the production value of these reference films.

## AEVRA music alignment check
The selected late excerpt starts at 185.333333 seconds. Low-frequency spectral-flux maxima within ±120 ms of each planned pulse have a median absolute grid offset of 13 ms across 43 anchors (less than one 24 fps frame). This checks approximate rhythmic alignment, not musical downbeat identification or subjective feel. Exact measurement data is included in `music-grid.json`.