# Camera arc: three briefs and editor handoff

Executed in Aixio on October 6, 2026. This is a manual sequence with an audited private saved Skill, not an automatic batch or guaranteed reconstruction.

## Choose the job

- Product concept: lamp card after the static hero, before a real finish board. Preserve shade rim, one stem and base. CTA: Review finish options.
- Editorial: frozen cup moment after a title, before ingredient details. Preserve cup, handle, lemon and all visible droplets. CTA: Explore the visual recipe.
- Architecture concept: pavilion opener before plans/materials. Preserve actual roof, posts, entrance and plinth. CTA: View concept plans.

Use an arc for depth-dependent overlap. Use a pan for reframing, a zoom for a detail, additional real views for factual product/property geometry. All subjects here are fictional concepts.

## Inputs and generation

1. Review the entire image/video plan, sufficient balance and one diagnosed correction allowance before any submission. Use current quotes; displayed rounding may differ from the ledger.
2. Generate a clear original source: Seedream 5.0 Pro, 2K, auto aspect, one image. Inspect immutable details, margins and depth landmarks. Our cup missed droplet count, lamp has a leaning stem, and pavilion has a pitched roof despite a flat-roof request. Preserve the accepted source or repair it when the missed detail is mandatory.
3. Add a video generator and attach its own accepted image via Attach. Lamp/pavilion used Your files; cup used the Canvas image. Confirm the source, not merely a node label.
4. Select Seedance 2.0 Mini; set 720p, four seconds, audio off. Do not save these as the owner's global defaults.
5. Write a subject-specific brief plus explicit camera instruction. Exact submitted text, including failed first pavilion and its correction, is in prompts.json.
6. Check the live quote, submit once, await completion. Verify admission before trying a click again. Download the original, inspect full playback and sampled frames.
7. Judge parallax, complete framing, stable subject geometry, no cuts and no scene action. Diagnose one cause before the single bounded correction. Stop when the reserve is used.

## Saved Skill audit

Private Skill: Small camera arc · rigid scene · article test, version 1. Reused on the lamp composer and through a connected Skill node for the pavilion. Both records retain the same Skill and correct own input. Lamp's submitted text explicitly contains the common camera instruction; the first pavilion contains only its brief. A connection alone is not evidence of instruction delivery. The first pavilion failed complete-roof and constant-scale checks and received one explicit correction. The correction still clips its opening and does not establish constant scale; only its last two seconds are accepted and held in the edit. Do not promise automatic reuse success.

Source acceptance, attachment, settings, quote review, playback judgment and editing remain manual. Frame sampling and local export can be automated after download.

## Editor handoff

Each selected output has an eight-second silent H.264, 24 fps, fast-start edit: title still 0–2 seconds, motion section 2–6 seconds, CTA still 6–8 seconds. Lamp/cup use four raw seconds. Pavilion uses raw 2–4 seconds at edit 2–4, then a two-second end-frame hold; its clipped opening is excluded. Source stills fit the native video aspect without stretching. Captions are added after generation. Edits are downloadable from the article media links; use the raw clips for alternate timing.

- lamp-sequence.mp4: square product concept. Follow with a real finish board.
- editorial-sequence.mp4: portrait frozen-moment editorial card. Follow with a recipe or ingredient detail.
- pavilion-sequence.mp4: landscape concept opener. Follow with the actual concept plans/materials, not an invented real-property tour.

Replace branding/CTA as needed, keep fictional-concept disclosure, supply a real destination, and review new crops on mobile. No sound was generated, no seamless loop is claimed. Preserve source, raw output, inspection frames and exact prompt alongside the delivery file.

## Cost record

Three images at 1.170 plus four videos at 3.931 = 19.234 credits, seven paid submissions, exactly one pavilion correction. New two-case expansion plus correction: 14.133 credits. No purchases or additional retries. This is a dated observed ledger, not a future price promise.

## Evidence limits

Modest creative viewpoint changes do not establish a measured trajectory, immutable world coordinates, hidden geometry or success rate. Full 360 LoRA and Meridian are separate documented methods, not executed Aixio settings. Generic troubleshooting and creator-comment fixes remain hypotheses unless this article marks them as an actual run.