Mininglamp_Product @mininglamp.com

Let "Meow" Take Over

「Meow」is an trending AI tool focused on efficiently handling ad creatives, balancing creativity and efficiency: not just analyzing, but also creating.

「Meow」tackles key pain points throughout the entire production process, effortlessly generate massive viral-ready short video creatives.

Why "Meow"?

Handle entire workflow with「Meow」

Step1 Viral ad video analysis

"Meow" delivers the best-performing creative ad videos for your products each day, delivers video breakdown ready-to-use—from visuals and scripts to structuring and pacing—accurately identify key content features. Notably, we've specially trained an AI with a keen sense for online trends. It does not just understand text; it also comprehends visuals, identifies golden 3 seconds, and extracts winning content strategies.

Step2_1 Script & Video Generation

"Meow" doesn't just analyze—it creates. Upload your product details, and the AI will automatically match viral content formulas to generate high-quality scripts. Once the script is ready, "Meow" assembles the final video using real footage. If no suitable clips are available, it can seamlessly generates and integrates AI-powered visuals, then adds voiceovers and subtitles—finalizing your video in minutes with only minor adjustments needed.

Step2_2 AI-Powered Asset Tagging

Organize scattered video assets with AI auto-tagging, enabling instant retrieval.

Based on self-developed model by Mininglamp Technology

Trending Large Multimodal Model
(HMLLM)*

In-depth interpretation of massive short advertising videos, extract effective content formulas.

We proposed the first evaluation benchmark for large multimodal exceeding 3 modalities in this field.

Subjective Response Indicator Advertisement Video (SRI-ADV) dataset, which is a large-scale data evaluation benchmark, we captured real-time electroencephalogram (EEG) and eye-tracking data generated by diverse populations when watching advertisement videos.

We proposed a brand-new large multimodal model algorithm paradigm

Hypergraph Multimodal Large Language Model (HMLLM) utilizes hypergraphs to construct complex relationships among video elements, EEG signals, and eye-tracking data. Through integrating information from different modalities, HMLLM bridges the semantic gaps between rich modalities, enhancing its logical reasoning and semantic analysis capabilities.

ACMMM2024 (CCF-A) conference altogether received 4385 effective submissions, among these, 1149 papers were accepted, 174 papers were selected as Oral, only 26 papers obtained Best Paper nominations (Mininglamp Technology's submission positioned 8th). ACMMM conference is a top international academic conference in the multimedia field hosted by the Association for Computing Machinery (ACM), it is also a China Computer Federation recommended A-class international academic conference (CCF-A). Conference topics cover all aspects of multimedia computing, such as multimedia content analysis, multimedia retrieval, multimedia security, human-computer interaction, and computer vision.

Featured cases