Built for performance marketers, Shopify DTC founders, and Amazon FBA catalog teams. Xiaomie Magic Carousel Video uses multimodal Vision AI to inspect your actual studio product shots, writes scroll-stopping hooks in 30 languages, pairs them with lifelike Microsoft Neural voiceovers and dual captions (.srt + burned-in), and renders 1080×1920 H.264 MP4s locally in your browser.
“Upgrade your desk ergonomics in 3 seconds — solid CNC aluminum!”
Compress photo sorting, foreign copywriting, voice recording, caption syncing, transitions, and BGM deduplication into a single folder click.
Built-in presets for AGNES 2.5 Flash and InternVL2.5-28B multimodal vision models. Automatically samples and compresses the first 4 photos of each product folder to identify materials, features, and use cases, generating high-converting native scripts validated across 30 language scripts.
Full support for 30 global languages (English, Spanish, Portuguese, Japanese, Korean, French, German, Russian, Arabic, Thai, Vietnamese, Indonesian, Hindi, etc.) with dedicated female and male neural voices. Automatically times sentences for both burned-in video captions and standalone .srt files.
Powered by FFmpeg.wasm directly inside your browser memory. Cropping, blurred background padding, dynamic zoom, transitions, and H.264 encoding happen on your own machine—keeping unreleased product photos confidential and eliminating server queue times.
Mix and match 10 transitions (fade, slideleft, slideright, slideup, slidedown, wipeleft, wiperight, circleopen, circleclose, pixelize) with smooth Ken Burns zoom. Background music automatically rotates through pitch and EQ micro-variants per product to reduce platform duplicate-content flags.
Select one root folder containing dozens of product subfolders and generate one video per subfolder automatically. Folders with fewer than 6 photos are cleanly logged and skipped individually without blocking or aborting the rest of your batch.
Beyond generating videos in 30 languages, the studio UI itself supports over 100 languages. On first visit, app titles, buttons, and instructions automatically match your IP country (defaulting to English where needed), with instant manual language switching anytime.
Performance teams need dozens of fresh 9:16 variations per week—without paying $150 per UGC clip or waiting on agency turnarounds.
Instead of generic slideshow text, Vision AI inspects textures, ports, stitching, and packaging from your first 4 product shots to craft problem-solution hooks tailored for US/UK shoppers.
Point the browser studio at your master SKU directory. Subfolders with 6+ photos automatically render into 1080×1920 H.264 MP4s with .srt closed captions for accessibility and video SEO.
Every frame is composited inside your local Chrome/Edge memory via FFmpeg.wasm. Your clients’ unreleased hardware or seasonal lookbooks never sit on third-party video servers.
Whether you are testing an Ergonomic Aluminum Laptop Riser in the US/UK or expanding a Cordless Pour-Over Coffee Grinder into Germany, Japan, and Mexico, one master folder generates localized voiceover creatives ready for immediate ad manager upload.
No video editing experience or heavy desktop installer required. Run a complete multi-language video production line right in Chrome or Edge.
Organize products as subfolders inside a master folder (each subfolder name becomes the product title, containing 6+ product photos). Click "Select Master Folder" to load the entire batch at once.
📁 Shopify_Q4_Drops/ ├─ 📂 Aluminum_Stand_US409/ (9 photos) ✅ ├─ 📂 PourOver_Grinder_US410/ (7 photos) ✅ └─ 📂 Embargo_Sample_US411/ (4 photos) ⏭️ Smart Skip
Pick from 30 voiceover/subtitle languages, select female or male voice, choose a copywriting angle (Natural Sales, Pain-Point Hook, Gift Scene, Minimal Luxury, or Promo Urgency), and toggle your preferred transitions.
🌐 Language: English / Español / 日本語 ... 🎙️ Voice: Female (Jenny) / Male (Guy) 🎬 Effects: Fade + Slide + CircleOpen + Zoom
Click "Start Batch Generation". Vision AI inspects the product images to craft a native script, Cloudflare Edge TTS synthesizes crystal-clear 24kHz audio, and subtitle timestamps are calculated automatically.
👁️ Vision AI → 📝 Native Sales Script 🔊 Sec-MS-GEC Edge Neural TTS (24kHz MP3) 💬 Auto-synced SRT Subtitle Timeline
FFmpeg.wasm combines photos, transitions, zoom, burned-in captions, voiceover, and deduplicated BGM into a 1080×1920 H.264 MP4, saving both the video and .srt file directly to your output folder.
🎬 Aluminum_Stand_US409_en.mp4 (1080×1920 H.264) 📄 Aluminum_Stand_US409_en.srt (Standard Subtitles) ✨ Proceeds to next product automatically
Scale localized vertical video assets across global storefronts and social channels in minutes.
Turn one set of product photos into localized vertical videos for US/UK, Latin America (ES/PT), Japan, Europe, Middle East, and Southeast Asia—complete with varied transitions and BGM adjustments.
Have hundreds of SKUs with photos but no listing video? Point Xiaomie at your SKU photo archive and batch-generate narrated product showcase videos overnight to boost listing conversion.
Transform static PDF catalogs into engaging 15–30 second narrated product clips in English, Spanish, Arabic, or Russian—ready to send to overseas buyers via WhatsApp or email.
Every batch exports both 1080×1920 MP4s and standard .srt subtitle files, enabling accessible captions and indexable multilingual video metadata across web storefronts and social platforms.
Built specifically to solve foreign copywriting, authentic voiceovers, platform deduplication, and unattended batch reliability.
| Capability | Manual Video Editing | Basic Slideshow Makers | Xiaomie Magic Carousel Video |
|---|---|---|---|
| Foreign Sales Script | Requires manual translation & copywriting per SKU | No script or static boilerplate text only | Vision AI inspects product photos to write 30-lang copy |
| Voiceover & Subtitles | Expensive voice actors or manual caption timing | Robotic voices, no .srt subtitle export | 30-lang Microsoft Neural F/M voices + burned & .srt captions |
| Platform Deduplication | Manual filter/audio tweaks for every export | Identical transitions & identical BGM loop | 10 randomized transitions + Ken Burns zoom + BGM pitch/EQ variants |
| Batch Fault Tolerance | One project at a time, no folder automation | Crashes or locks batch if a folder has <6 photos | Subfolder batch automation; cleanly skips <6 photo folders |
| Privacy & Setup | Multi-GB desktop software & heavy GPU install | Uploads confidential product photos to cloud queues | Zero-install browser WebAssembly; photos stay on your PC |
The browser studio is live now. Bring your own Vision API key or connect with our team for enterprise relay & localized payment plans.
Open carousel.xiaomie.tools in desktop Chrome or Edge to use the full batch synthesis pipeline immediately, including local folder I/O, 10 transitions, BGM deduplication, and 30-language Edge TTS.
For high-volume seller teams needing dedicated private TTS/Vision relays, centralized quota management, or custom branding, with multi-currency and local payment methods supported.
Everything you need to know about image thresholds, AI models, voiceover synthesis, and data privacy.
Six or more images ensure smooth pacing, visual variety, and strong retention in a 9:16 short video. During batch processing, if any product subfolder has fewer than 6 images, Xiaomie logs the reason and skips only that subfolder—never locking or stopping the rest of your batch.
Video rendering happens 100% locally inside your browser using FFmpeg.wasm, and MP4/.srt files save directly to your disk. Only when AI Vision Copywriting is enabled are the first 4 images compressed into lightweight thumbnails and sent to your chosen Vision API (AGNES 2.5 Flash or InternVL2.5-28B) to generate the script.
Video generation (AI copywriting + Neural TTS + subtitles) supports 30 languages including English, Simplified/Traditional Chinese, Japanese, Korean, French, German, Spanish, Portuguese, Russian, Arabic, Hindi, Indonesian, Vietnamese, Thai, Turkish, Italian, Dutch, Polish, Ukrainian, Swedish, Czech, Danish, Finnish, Greek, Hebrew, Hungarian, Norwegian, Romanian, and Persian. The software UI itself supports 30 languages, automatically selected by visitor IP with manual override.
Yes. Subtitles and .srt files are generated whenever "Burn Subtitles" is enabled, timed evenly across the video duration, and background music with pitch/EQ anti-duplication is mixed independently of TTS.
We recommend the latest desktop version of Microsoft Edge or Google Chrome, which support the File System Access API (for one-click master folder reading and automatic file saving) and high-speed WebAssembly execution.