- Home
- Skills
- Image & Video
- Beat Drop Reels: Beat-Synced 9:16 Promo Videos From Photos
More screenshots
Works with the AI Agents you already use
Beat Drop Reels: Beat-Synced 9:16 Promo Videos From Photos
Turn product photos, a logo and a song into a finished 1080x1920 promo MP4 for Reels, TikTok and Shorts: every cut lands on the beat, verified by a built-in sync check.
$15
Beat Drop Reels: Beat-Synced 9:16 Promo Videos From Photos
Example session with this skill installed
We're doing a one-day pop-up for MOTHWAVE at Dock 9 Market on Saturday 14 November, 12-6pm. Our logo, product cut-outs (Drifter 01 sneaker in Ember and Coal, the Static hoodie in black and sand, Late Night Roast coffee) and our own track beat_124.wav are in ./mothwave-assets. Can you make a 20-second vertical reel for Instagram and TikTok that hits the beat? Free entry, the first 50 people get a free coffee, map is the link in bio. Keep it on-brand: black and our orange #FF7A1A.
- Read your context and instructions
- Compiled the beat drop reels
- Generated the image
Your pop-up reel is ready: reels/mothwave_popup_20s.mp4
Specs: 20.0 s, 1080x1920, 30 fps, H.264 High (yuv420p) + AAC, loudness -14.2 LUFS, faststart, 30 MB.
qa.py check: PASS on all 11 checks.
Beat sync: your track is 124 BPM and the drop is at 7.74 s; the reel starts 4 beats before it, so the
loading bar hits 100% exactly on the drop. qa.py sync measured 48 visual hits after the drop: 37 on a beat,
11 on a half-beat, 0 off the grid, and all 32 planned cuts were found. Median offset +0.5 ms, worst 16 ms
(half a frame). I can't listen to audio, so sync was verified from the MP4's own soundtrack and from your
original WAV, plus the frames below.
What happens when
| Time | On screen |
|---|---|
| 0-1.9 s | Build: logo, ONE-DAY POP-UP, SAT 14 NOV badge, loading bar "OPENING DOORS..." |
| 1.9-3.4 s | Drop: POP-UP / THIS / SATURDAY, one word per beat (SATURDAY in orange) |
| 3.4-5.3 s | Details pop on each beat: SAT 14 NOV, 12-6PM, DOCK 9 MARKET, FREE ENTRY stamp |
| 5.3-8.2 s | Drifter 01, "Try them on in person", Ember and Coal swapping on every half beat |
| 8.2-11.1 s | Static Hoodie, Black and Sand |
| 11.1-13.1 s | Late Night Roast, "First 50 in: free coffee" |
| 13.1-14.0 s | THE WHOLE DROP / IN ONE ROOM lineup |
| 14.0-20.0 s | End card: SAT 14 NOV, 12-6PM / DOCK 9 MARKET, FREE ENTRY, FIRST 50 IN GET A FREE COFFEE, MAP: LINK IN BIO, logo; fades to black so the loop restarts cleanly |
I checked the encoded frames at the drop, each product and the end card: everything is readable and sits
inside the safe area, clear of the TikTok/Reels buttons and caption zone.
Caption
One day only: Saturday 14 Nov, 12-6PM at Dock 9 Market. Try the whole drop on in person. Free entry, and the
first 50 through the door get a free coffee. Map in bio.
#mothwave #popupshop #streetwear #sneakers #dock9market
Notes
- No prices are on screen because you didn't give any. Send them and I'll add a price tag per colourway and re-render (about 30 seconds).
- The music is your own track, so there are no rights issues. Upload this file as-is. If you need a copy under 9 MB for WhatsApp or email,
qa.py compresswill make one. - Before posting, make sure the link in your bio points to the map.
- Also saved: reels/mothwave_popup_20s_stills.jpg (QA contact sheet) and reels/mothwave_popup_20s_plan.json (beat map).
mothwave_popup_20s_stills.jpg
PNG · 1536×1024
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Most "make me a promo video" prompts end in a storyboard. This skill ends in an MP4, with proof that every cut hits the beat.
Give your agent a logo, a few product images and a song. It maps the music (tempo, beat grid, the drop), writes a short scene config, checks a contact sheet of stills, renders a 1080x1920 vertical video frame by frame in Python, then checks the encoded file: codec, loudness, faststart, and where every visual cut lands against the beat. You get a ready-to-post MP4, a caption and a timing report.
What you get
- 4 ready reel formats: product drop / launch, SOLD OUT -> RESTOCKED (the stamp flips on the beat), countdown with a silhouette teaser and reveal, and event / pop-up promo. 7 scene types (intro, hook, product, lineup, countdown, card, end) cover showcases, sales and single-product heroes too.
- 3 looks from one switch:
bold(streetwear and merch: outlined titles, grain, glitch hits),clean(premium DTC: soft shadows, rounded cards, spotlights, whip transitions),neon(nightlife and events: glowing type, RGB split). - Beat-locked motion: the scene you choose starts exactly on the drop. Hook words slam one per beat, products swap on half beats, price tags and stamps pop on the beat, a loading bar hits 100% on the drop, with synth impacts and whooshes on the cuts.
- Platform-ready output: H.264 High, yuv420p, 30 fps, AAC, loudness-normalised to -14 LUFS, faststart, capped at about 12 Mbps, plus a two-pass compressor for size limits (for example under 9 MB).
- Safe-zone aware: key text stays clear of app buttons and captions, and
stills --safetints the covered areas so you can check before rendering.
Why it beats prompting a video model or an editor template
| The usual problem | What this skill does |
|---|---|
| Cuts "roughly" on the beat, drifting off by the end | Tempo and phase are fitted to the song; cuts are placed on the beat grid and land on the nearest frame (within half a frame) |
| "It's in sync" is a claim | qa.py sync re-measures the beat from the finished MP4's soundtrack and from the original song, finds every cut in the picture, and reports the offsets. On the 4 demo reels: 0 cuts off the grid, median offset within 5 ms, worst 41 ms (just over one frame at 30 fps) |
| AI video invents your product | Uses your own images: transparent cut-outs float with shadows and halos, photos become cards. Nothing is generated or hallucinated |
| Files that fail upload or sound quiet | qa.py check verifies codec, profile, pixel format, frame count, duration, AAC, -14 LUFS and faststart, and prints PASS |
| Templates locked inside an app | Plain JSON config plus Python, so every edit re-renders reproducibly in about 30 seconds |
What's inside
SKILL.md: a 9-step workflow (preflight, brief, asset check, music map, plan, stills, render, QA, hand-over), reel recipes, a quality bar, troubleshooting and a hand-over template- 6 Python scripts:
reel.py(renderer),reelkit.py(engine),analyze_audio.py(BPM, beat grid, drop finder, drum hits),qa.py(check, sync, frames, strip, compress, sheet),preflight.py(dependency check with install commands),make_demo_assets.py(an original demo kit: 3 fictional products, a logo and 2 synthesised beats, so you can try it before your assets are ready) - References: full config schema, a creative guide (hooks, CTAs, safe zones, captions) and a cookbook for custom effects
- 4 working example configs
Who it's for
Merch and streetwear brands, Shopify and DTC stores, coffee, beauty and food brands, musicians selling merch, event organisers and pop-ups, and social media managers and freelancers who make launch, sold-out, restock and countdown posts every week and want them on the beat without opening an editor.
Requirements
An agent that can run code and Python (tested in Claude Code), with Python 3.8+, Pillow, numpy and ffmpeg. preflight.py checks all of them and prints the exact install command for Windows, macOS or Linux. No API keys, no network.
- Motion graphics built from still images: no footage editing, no 3D, no AI-generated imagery. Product cut-outs (transparent PNGs) look best; plain photos are shown as cards.
- Beat detection is strongest on music with a clear pulse (pop, hip-hop, electronic, rock). For live, rubato or ambient tracks, set bpm and drop by hand (analyze_audio.py --hits lists the real drum hits).
- Output is 1080x1920 at 30 fps only (no 4:5 or 16:9 version yet).
- Fonts come from your system or a fonts/ folder. Brand fonts are not included; without good fonts it falls back to system faces such as Impact, Arial Black or DejaVu.
- The agent cannot hear the result: sync is verified numerically from the waveform and by looking at extracted frames.
- Needs a local Python + ffmpeg setup, so chat-only agents without code execution can't render.
- You need the rights to the music, images and fonts you use. Copyrighted songs can be muted by platforms.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 month ago
Requires
Permissions
File Scopes
Runs the bundled Python scripts locally and calls ffmpeg/ffprobe (as argument lists, never a shell) to analyse your song and encode the MP4. Reads only the images, logo and song you point it at; writes renders, stills and QA logs into your output folder. No network access, no environment variables, no secrets.