Make motion
with ChatGPT.

Write a Malayalam Instagram Reel script for [AUDIENCE] about [TOPIC].
Target about [DURATION] seconds, then revise timing after recording the voice.
Give 3 opening hooks that create curiosity without exaggerated claims.
Choose one and write: question or result first, immediate payoff, 3 useful beats,
and one clear CTA. Use natural spoken Malayalam with familiar English terms.
Return a separate clean TTS script without scene labels or stage directions.
Do not claim the final video can be rendered unless the tools are available.

Turn the approved script into a timed visual storyboard.
For each beat give: approximate time, spoken phrase, short on-screen text,
main graphic, entrance, exit, transition and sound cue.
The first frame must already show the hook. Show the result before explaining.
Use 5-7 distinct visual beats. Give every motion a purpose.
Keep important text inside the Instagram-safe content area.
Mark timing as provisional until the narration is supplied.

Use the HyperFrames skill to build an editable Malayalam Instagram Reel.
Subject: [TOPIC]. Audience: [AUDIENCE]. CTA: [CTA]. Handle: [HANDLE].
Format: 1080 x 1920, 30 fps. Target duration: [DURATION] seconds.
Read the project instructions and inspect my attached reference first.
Use the approved script and narration. Let the actual voice set scene timing.

Creative direction:
Near-black deep green, acid lime, warm white; restrained violet/teal accents.
Use large Malayalam headlines, tilted lime phrase plates, dimensional portrait
video cards, one contrasting white phone scene, code-to-preview motion and a
finished-Reel payoff. Show the result in the opening. Avoid generic slide decks.
Fill the composition with useful foreground art and restrained background detail.
Keep key copy between y=180 and y=1330, clear of the right action rail.
Use local Malayalam fonts and verify shaping; keep reading text at least 28 px.

Motion direction:
Mix masked type reveals, perspective card entrances, a purposeful tool-card flip,
prompt typing, staged UI responses and a clear export payoff.
Use overlapping entrances and varied eases. Land titles and let viewers read.
Use real supplied brand assets and product captures where needed. Label any
illustrative workflow UI as a preview; do not pass it off as a real screenshot.
Add licensed action sounds: whoosh on major movement, pop on a card landing,
keys on typing, one impact on the main payoff. Keep voice intelligible.

Implementation and delivery:
Use one scene per editable sub-composition and one paused seekable GSAP timeline
per composition. Keep media/font assets local and use deterministic animation.
Start with a storyboard, then build the actual project files.
Run the full HyperFrames check, inspect phone-size frames and every scene seam.
Fix clipped text, collisions, empty initial frames, missing media and dead stretches.
Open a persistent browser preview with sound. Show me that preview for review.
After I approve it, render the MP4, verify its dimensions/duration/audio, and give
me the actual output file plus the editable source. Do not stop at a code snippet.

Use assets/audio/narration.mp3 as the timing source.
Measure its duration. Find the actual phrase starts and align the hook, answer,
tool change, prompt, animation, export and CTA to those spoken phrases.
If transcription translates Malayalam to English, use it only for timing.
Keep the visible captions in accurate Malayalam and check them against the audio.
Do not shorten, speed up or regenerate my voice without asking.
Let the final CTA hold for about 2 seconds after it becomes readable.
Check that no word or sound is cut off at the end.

Make a cue list with each sound file, start time, duration, gain and fade.
Use supplied or licensed sounds; do not invent a license or a downloaded asset.
Align sound peaks to the actual visual landings, not just clip start times.
Raise quiet narration gently and keep the voice ahead of the effects.
Measure the combined mix, not only individual files. As a starting target for
this Reel, aim around -16 to -14 LUFS integrated with true peak at or below
-1 dBTP. Adjust by listening; these are production targets, not platform rules.
Prevent clipping when impacts and voice overlap. Keep the source files and cue
list editable. Show the measured levels and tell me what was not auditioned.

At [TIME] the [ELEMENT] feels [SPECIFIC PROBLEM].
Change it to [CONCRETE ACTION], and land it on the spoken phrase [PHRASE].
Example: at 0-1s, make the question readable from frame zero. Increase headline
size, fan three portrait result cards into the center, then hold the title still.
At export completion, reveal the actual finished Reel thumbnail instead of
ending with a progress bar alone. Fix only the affected timing/layout, then
inspect those frames and run the required check again.

Create a portrait 9:16 Instagram Reel cover for [HANDLE].
Use near-black deep green, acid lime and warm white, with a polished dimensional
video/play graphic and restrained purple/teal accents.
Keep the headline large and all essential content within the central crop area.
Exact text: [HEADLINE]. Handle: [HANDLE].
Use correctly shaped Malayalam lettering. Avoid extra copy and watermarks.
Give me the final image file, then inspect spelling and a center-crop preview.