Create an editable assistant conversation launch video for [PROJECT] and [AUDIENCE]. Build a new result from [APPROVED_CONTENT] and [BRAND_RULES], using the credited reference to discuss composition and timing.

## Required Content and Inputs

Introduce a small assistant avatar and one user need; type a concrete request; reveal the supported response in cards; show a follow-up action and final identity.

Ask for exact on-screen copy, authorized assets, intended destination, output dimensions and licensed audio. Identify missing task data before implementation. Keep real claims and sample data distinct; do not invent customer results or capabilities.

## Scene Plan

1. Introduce a small assistant avatar and one user need. Define the visible action, its result and the transition anchor.
2. type a concrete request. Define the visible action, its result and the transition anchor.
3. reveal the supported response in cards. Define the visible action, its result and the transition anchor.
4. show a follow-up action and final identity. Define the visible action, its result and the transition anchor.

Return a timing table with scene start and end, copy, assets, motion and audio cues. The reference is 82.6 seconds at 1752 × 1080; adapt the timeline to the supplied content instead of forcing unreadable text into that duration.

## Motion and Visual System

Use one conversation record for the prompt, typing state, response and result cards. Give the response enough reading time and keep the mascot secondary to the task.

Use the selected project design system. Establish final readable states before animating entrances. Preserve image and logo proportions, separate editable type from raster art and use a restrained hierarchy for supporting labels. Keep one recognizable anchor between adjacent scenes.

## Implementation and Audio

Use the existing project stack and an explicit timeline that can be inspected at any timestamp. Store copy, data, colors, scene timing and motion parameters in editable configuration. Seed any randomness so the same timestamp renders consistently. Do not depend on wall-clock timers to define an export frame.

Map supplied music or sound cues to timeline markers. If audio is missing, deliver a silent timing preview first. Do not claim beat or voice synchronization without the actual authorized track.

## Review Priorities

- Verify every displayed capability with supplied product evidence. Label scripted output as a demonstration and never present invented personal data as a real customer session.

Inspect the longest real label, the first and last frame and both sides of each scene boundary. Check phone-size reading, foreground occlusion and the actual output aspect ratio. Compare the exported duration, frame count and audio tail with the approved timeline.

## Deliverables

Return editable source, the timing table, one still per scene, a poster and an MP4. Provide the low-cost scene preview before the full render. Distinguish a pre-rendered interface demonstration from working application behavior.

Reference creator: @trymirage.
Source: https://x.com/trymirage/status/2102863208209932736