The article examines common pitfalls of naïve ad insertion in short‑video feeds, explains how cursor‑based pagination prevents duplicate ads, and outlines a client‑side mixing architecture that isolates services, meets strict latency requirements, and ensures accurate ad billing.
This article dissects the end‑to‑end architecture of a billion‑scale short‑video platform, detailing layered design, core services such as upload, transcoding, recommendation, interaction, storage, and the key challenges of massive video storage, high‑concurrency streaming, low‑latency playback, and real‑time recommendation reliability.
MoneyPrinterTurbo, an open‑source Python project, uses multiple large‑model APIs to automatically generate scripts, fetch royalty‑free footage, synthesize speech, add subtitles and music, and render HD short videos with a single click, targeting creators, marketers, SMEs, developers, and educators.
This guide breaks down a step‑by‑step workflow—starting with tool selection, character design, and script drafting, then moving through batch asset generation, automated video creation, transition tricks, and final polishing—to help beginners produce a high‑impact AI‑generated short drama within three days.
The article presents a comprehensive guide to building a free open‑source short‑drama mobile app, detailing functional modules, immersive player, personalized recommendations, social interactions, monetization models, and a full tech stack—including Flutter or React Native for the front‑end, micro‑services backend, CDN, and DRM—while addressing high‑concurrency and cross‑platform challenges.
The article outlines a complete open‑source architecture for a short‑video platform, detailing front‑end player features, personalized recommendation, multi‑mode playback, social interactions, monetization models, cross‑platform technology choices, backend micro‑services, CDN, caching, security, and solutions to high‑concurrency and synchronization challenges.
This article dissects the architecture of a modern short‑video backend, covering layered system design, core services such as video production, distribution, interaction, storage strategies, real‑time and offline recommendation engines, high‑concurrency streaming solutions, and practical techniques for cost control, scalability, and fault tolerance.
This article analyzes KuaiMod, a multimodal large‑model solution developed by Kuaishou for short‑video content quality assessment, detailing its benchmark dataset, chain‑of‑thought data construction, offline SFT + DPO training, online reinforcement‑learning updates, evaluation results, and large‑scale deployment impact.
The second short-form video quality assessment and enhancement challenge, co‑organized by Kuaishou's audio‑video team and the Intelligent Media Computing Lab, invites global researchers to develop efficient quality assessment models and diffusion‑based super‑resolution methods using the new KwaiSR dataset, with prize money and potential CVPR workshop paper invitations.
This guide shows how the DeepSeek AI model combined with Jianying video editor can automate topic selection, script writing, voice‑over, editing and distribution, boosting short‑video creation efficiency by up to tenfold and enabling creators of any skill level to produce professional‑grade content at scale.
This article surveys recent advances in applying causal inference to recommender systems, presenting three lines of work—causal embedding for interest‑conformity disentanglement, contrastive learning for long‑term and short‑term interest separation, and adversarial debiasing of duration bias in short‑video recommendation—along with experimental validation and insights.
The CVPR NTIRE 2024 short-form video quality assessment competition introduced the KVQ dataset, attracted over 200 teams, evaluated submissions using SROCC and PLCC metrics, and highlighted the winning approaches of SJTU MMLab, IH‑VQA, and TVQE, showcasing advances in AI‑driven video quality evaluation.
This article introduces the open‑source Douyin‑Vue project—a TikTok‑style short‑video app built with Vue 3, Vite 5 and Pinia—provides its repository and demo links, and walks through the required environment setup and local deployment steps.
This article introduces the Douyin open‑source short‑video project built with Vue and Vite, provides a free book giveaway link, and gives step‑by‑step instructions for installing Node, cloning the repository, installing dependencies with pnpm, and running the application locally.
The article explains why first‑frame latency matters for video apps and presents a comprehensive set of optimization methods—including pre‑fetching URLs, intelligent preloading, prerendering, and scenario‑specific tweaks for cold start, short‑video scrolling, page navigation, and long‑video playback—to dramatically reduce start‑up time and improve user experience.
This talk presents Kuaishou's research on combining reinforcement learning with multi‑task recommendation, detailing a two‑stage constrained actor‑critic method for short‑video ranking, a multi‑task RL framework, experimental results on offline and online systems, and practical Q&A insights.
This article highlights three oral papers accepted at WSDM 2024 that address cross‑domain sequential recommendation, extremely sparse feedback denoising recommendation, and automated label crafting for short‑video recommendation, providing their abstracts, author lists, and links to PDFs and source code.
The CVPR NTIRE 2024 workshop hosts the first short‑video quality assessment competition, introducing the KVQ dataset of 4,200 videos across nine scenes, providing training/validation data, a baseline 3D Swin‑Transformer model, detailed competition rules, rewards, and organizer contacts.
The CCIR 2023 conference in Beijing, sponsored by Kuaishou, hosted a short‑video recommendation algorithm frontier research forum where over 100 experts and students shared the latest AI‑driven recommendation technologies, open datasets, and interdisciplinary challenges in short‑video platforms.
The Kuaishou Academic Forum held on November 23 in Beijing presented cutting‑edge research on short‑video recommendation algorithms, featuring talks on reinforcement learning, user interest modeling, graph neural networks, and a comprehensive recommender‑system simulator, while also offering registration details and a brief company overview.
Douyin has begun testing a paid short‑video model, offering the first five seconds for free and requiring users to purchase the rest with virtual Douyin coins ranging from 10 to 100, while creators must meet follower and verification criteria and the platform takes a 30% service fee.
The presentation introduces a two‑stage constrained actor‑critic algorithm that learns auxiliary policies for interaction signals before optimizing watch‑time under KL constraints, and a reinforcement‑learning multi‑task learning framework that models session‑level dynamics with adaptive multi‑critic weighting, both achieving significant offline and online gains in short‑video recommendation.
This article presents a two‑stage constrained actor‑critic reinforcement learning algorithm for short‑video recommendation, models the problem as a constrained MDP, details the algorithm’s stages, and reports extensive offline and online experiments showing superior watch‑time and interaction metrics, followed by a multi‑task RL framework and its evaluations.
The article highlights five CIKM 2023 papers covering a lightweight model‑compression framework for recommender systems, a query‑dominant user‑interest network for large‑scale search ranking, a causal watch‑time labeling approach for short‑video recommendation, implicit negative‑feedback optimization for short‑video feeds, and the KuaiSAR unified search‑and‑recommendation dataset, each with download links, author lists, and key findings.
This article introduces Kuaipedia, Kuaishou's multimodal short‑video encyclopedia, detailing its background, system architecture, knowledge‑video recognition pipeline, multimodal entity linking techniques, and downstream applications, while also providing implementation insights and a brief Q&A.
This article explains how to build a comprehensive user incentive framework for a short‑video platform, covering the definition of incentive systems, three core incentive types, their role across acquisition to monetization stages, and a detailed, data‑driven redesign for Haokan Video’s user segments.
The paper models short‑video recommendation as a constrained Markov decision process and introduces a two‑stage constrained actor‑critic algorithm that jointly maximizes watch time while satisfying multiple interaction constraints, demonstrating superior offline and online performance on the KuaiRand dataset and Kuaishou app.
This paper presents RLUR, a reinforcement‑learning algorithm that models user‑retention optimization as an infinite‑horizon request‑based Markov Decision Process, addressing uncertainty, bias, and delayed reward challenges to directly improve retention, DAU, and engagement in short‑video recommendation platforms.
On January 25, Kuaishou’s community science team announced that seven of its papers were accepted at the ACM Web Conference 2023 (WWW’23), covering reinforcement‑learning‑based user retention, constrained actor‑critic recommendation, divide‑and‑conquer embedding retrieval, causal embedding with contrastive learning, latent action space exploration, dual‑interest factorization attention, and multi‑task reinforcement learning for recommendation.
The Kuaishou audio‑video team’s PDAS algorithm, a probability‑driven adaptive streaming solution for short videos, achieved first place at ACM MM 2022 by modeling user behavior, optimizing pre‑load decisions, and delivering superior QoE while minimizing bandwidth waste.
This article summarizes four CIKM 2022 papers that present a client‑side short‑video recommender, the fully‑observed KuaiRec dataset, the unbiased KuaiRand sequential recommendation dataset, and an industrial‑scale solution for billion‑user lifetime value prediction, highlighting their motivations, methods, and reported impacts.
The article outlines a low‑cost, controllable face‑style transfer pipeline that preserves identity while applying customizable artistic styles, enabling immersive live‑stream and short‑video experiences, enhancing visual appeal, protecting privacy in buyer‑show images, and supporting scalable deployment across diverse faces, angles, and devices.
This article explains how Kuaishou detects and predicts viral short‑video素材 by defining content types, outlining essential viral elements, describing a two‑stage coarse‑recall and fine‑ranking model that combines speed‑based features, Gaussian mixture modeling, and a lightweight DNN, and showcases real‑world case studies and Q&A.
This article explains the background, design principles, high‑concurrency handling, storage optimization, rule‑engine implementation, and intelligent scheduling needed to create a universal, stable, extensible, and intelligent recommendation slot that enriches short‑video app ecosystems.
Meituan leverages computer‑vision techniques to tag, analyze, and automatically generate short videos across consumer and merchant scenarios, detailing hierarchical tag design, self‑supervised representation learning, fine‑grained food recognition, intelligent cover creation, and pixel‑level editing to enhance content discovery and presentation.
This article presents Kuaishou's graph‑based item‑to‑item (I2I) recall pipeline for short‑video recommendation, detailing the business challenges, pipeline architecture, optimization techniques such as similarity‑measure tricks, graph structure learning, edge‑weight learning, and future research directions.
This article explains how Kuaishou leverages graph neural networks for item‑to‑item (I2I) recall in short‑video recommendation, detailing the system background, pipeline architecture, optimization techniques such as similarity measurement, graph structure learning, edge‑weight learning, and future research directions.
This article presents Kuaishou's graph‑based item‑to‑item (I2I) recall pipeline for short‑video recommendation, detailing the business challenges, the advantages of graph neural networks, the system architecture, optimization techniques such as similarity metric refinement, graph structure learning, edge‑weight learning, and future research directions.
The framework tags short videos by fusing text, image and audio‑video features through specialized extraction, classification, generative and retrieval modules, then ranking candidates with a multimodal BERT model, delivering accurate, business‑specific tags that boost recommendation, search and advertising.
iQIYI’s short‑video team solves the new‑video cold‑start problem by using a GAN that generates latent user features from video attributes and a discriminator to validate them, then matches these vectors to real users via cosine similarity, achieving double‑digit gains in exposure, CTR, and watch time.
Kuaishou’s audio‑video team built a dedicated high‑quality transcoding tier for Jackie Chan’s debut on the platform, deploying AI‑driven analysis, enhancement and compression techniques to deliver ultra‑clear short videos while keeping cost and latency low, and describing the end‑to‑end processing pipeline and trigger mechanisms.
Kuashou’s audio‑video team built a custom short‑video transcoding tier that combines AI‑driven enhancement, aggressive compression and a specialized delivery strategy to meet billions of viewers’ high‑quality expectations for Jackie Chan’s debut, showcasing rapid cross‑team collaboration and advanced video processing techniques.
This article explains how to optimize Android short‑video playback for instant start by analyzing problems such as cover‑screen flash and slow streaming, then presents a three‑player RecyclerView solution, pre‑download techniques, cache integration, and future improvements to deliver a seamless, TikTok‑like experience.
CONDE, a concept‑aware denoising graph neural network proposed by Wuhan University and Kuaishou, leverages heterogeneous three‑part graphs, attention‑based graph convolutions, and Gumbel‑Softmax‑driven edge sampling to filter noisy user‑video interactions, achieving up to 6 % AUC improvement on short‑video and e‑commerce recommendation tasks.
This article examines the rapid growth of short‑video platforms, outlines the core problems a recommendation system must solve for users, creators, and advertisers, describes the end‑to‑end pipeline, explores multi‑objective modeling and fusion techniques, and discusses long‑term value estimation for sustained user engagement.
The article outlines Baidu Haokan’s short‑video recommendation architecture, describing how a unified ranking pipeline uses user‑interest signals, multi‑objective MMOE and deep‑fusion models, and long‑term value estimation to balance personalized user experience, creator exposure, and advertiser goals across billions of daily video plays.
This article analyzes Kuaishou's short‑video quality advantages by detailing its production‑consumption chain, multi‑dimensional evaluation methodology, objective and subjective test results, and key metrics such as bitrate savings, VMAF gains, win and selection rates, supported by visual examples.
This article reports an eye‑tracking study conducted on a short‑video app to uncover users' visual strategies, evaluate the impact of title and recommendation layout variations, and identify subtle usability issues that traditional methods miss, providing data‑driven guidance for product design decisions.
iQIYI improves short‑video recommendation by applying multi‑objective optimization—weighting clicks by watch duration, fusing separate click and watch‑time models, employing multi‑task learning with ESMM/MMOE and Pareto‑guided PSO hyper‑parameter search—delivering 7%+ watch‑time growth, 20%+ interaction gains, and 1.5‑3% CTR lifts while planning cross‑scene learning and further model refinements.
This article chronicles Kuaishou’s ten‑year journey from a humble GIF‑making app created by Cheng Yixiao, through its partnership with Su Hua, strategic pivots, IPO success and subsequent controversies, illustrating how focused product design and relentless R&D management built China’s premier short‑video platform.
This article chronicles Kuaishou's evolution from a modest GIF‑sharing tool to a market‑leading short‑video platform that listed on the Hong Kong Stock Exchange with a valuation exceeding HK$1.3 trillion, highlighting its funding rounds, product strategy, competition with Douyin, and aggressive growth campaigns.
The article recounts ByteDance’s 2016 decision to go all‑in on short‑video, analyzes the contrasting thinking models of its product manager, director, and founder Zhang Yiming, and explains how multi‑thinking, optimal‑solution, and super‑allocation mindsets drive strategic product success.
In the post‑traffic era, short‑video platforms must shift from relying on public‑domain traffic to leveraging private‑domain monetization, focusing on high‑quality creators, analyzing their needs across verticals, and building a research framework that guides platform policies, creator support, and ecosystem growth.
This article details Likee's short‑video recommendation pipeline, covering the evolution of its feature‑engineering framework, the use of Flink for minute‑level statistical and second‑level session features, the integration of automatic statistical features into DNN models, multimodal feature extraction, and the cache‑tensor technique that dramatically improves online inference performance.
This article examines how short‑video platforms can expand into new user circles by leveraging external trend data, competitive analysis, KOL resources, precise content matching, and data‑driven product strategies to acquire, retain, and reactivate users while balancing growth and platform health.
The article surveys iQIYI’s multimodal short‑video content‑tagging pipeline, detailing extraction‑ and generation‑based methods, challenges of open‑world tags, model evolution from rule‑based to Transformer generators, visual‑text fusion techniques, and applications such as recommendation, search, clustering, and future enhancements.
This article describes iQIYI's short‑video content tagging system, outlining the challenges of extracting type and abstract tags from multimodal data, detailing the evolution from text‑only models to image‑fusion, BERT‑enhanced, and video‑frame models, and discussing their applications and future directions.
This article examines the challenges of short‑video content tagging and describes iQIYI's multi‑stage evolution from simple text‑only models to sophisticated multimodal architectures that fuse cover images, BERT embeddings, and video frames to improve tag generation accuracy.
iQIYI’s short‑video tagging system tackles multimodal fusion, open‑set and abstract tags by evolving from a text‑only model through cover‑image, BERT‑vector, and video‑frame fusion architectures, enabling automated labeling, personalized recommendation, and semantic search while planning to add OCR, audio, and knowledge‑graph enhancements.
The paper proposes a Siamese‑based PredictionNet that combines EfficientB3 image and VGGish audio features with user metrics to predict a HotValue score for newly uploaded short videos, using a margin loss with view‑value‑aware pair selection, enabling tiered cold‑start exposure that boosts overall platform efficiency.
This article explains the rapid growth of short‑video on Taobao, describes the video feature framework, details the RankI2V and RankV2V recall methods, outlines coarse and fine ranking models, and presents real‑time interest and business strategies that significantly improved click‑through rates and viewing time.
Effective short‑video recommendation relies on meticulous feature engineering that transforms raw signals—numerical counts, categorical IDs, content and user embeddings, context and session data—through bucketization, scaling, crossing, and smoothing, then selects and evaluates them via filtering, wrapping, regularization, and importance analysis to mitigate business biases and improve multi‑objective ranking performance.
This article presents a comprehensive overview of short‑video recommendation at UC, covering business background, system architecture, the evolution from LR to Wide & Deep models, multi‑objective loss design with positive‑sample weighting, graph‑embedding fusion, time‑weighted loss, continuity modeling, a Boosting‑based WnD solution, and future research directions.
Youku’s MoKu Lab combines rule‑based preprocessing with multimodal AI models that score titles, covers and video content, enabling large‑scale machine filtering that boosts human review efficiency by 5‑6 percentage points and reduces low‑quality short videos from roughly 15 % to under 3 % while supporting flexible business rules.
This article explains how MaFengWo’s iOS app implements a flexible video editing framework that supports custom editing and template‑based creation, detailing the requirement analysis, architecture design, AVFoundation‑based implementation, performance optimizations, and future directions for short‑video content.
iQIYI’s large‑scale hierarchical classification system combines multimodal text and image embeddings, low‑rank multimodal fusion, and a dense hierarchical multilabel network with cascade‑style weighting to assign accurate type tags to short videos, boosting production efficiency and personalized recommendation diversity.
The article presents a detailed analysis of Weibo's short‑video platform architecture, covering team background, business scenarios, micro‑service design, feed‑pull model, multi‑level distributed caching, multi‑datacenter HA deployment, circuit‑breaker mechanisms, and elastic scaling to achieve high availability under unpredictable traffic spikes.
This article presents a comprehensive evaluation of AV1's compression efficiency and encoding complexity on short videos, comparing it with x264, x265, and VP9 using BD‑Rate, PSNR, SSIM, and VMAF‑Phone metrics across multiple resolutions and bitrate control modes.
The article outlines the technical challenges of short‑video apps and presents the modular, extensible architecture of the 58 Short Video SDK, detailing its layered design, design principles, advantages, and future evolution to support advanced features such as AR, hardware decoding, and h265 encoding.
This article details the evolution of Alibaba's short‑video feed ranking system, from a Wide&Deep CTR model to a hybrid Factorization‑Machine and Gradient‑Boosted‑Tree approach, describing feature engineering, model architecture, experimental results, lessons learned, and future directions toward duration‑based relevance.
In the 2018 AI Challenger, Meitu co‑organized a new “Real‑Time Short‑Video Classification” track and released the industry’s first multi‑label short‑video dataset of 200,000 mobile‑captured, vertically oriented videos spanning 63 categories and detailed tags for subjects, scenes, actions, and other dimensions, advancing video semantic understanding and AI research.
This article explains how short‑video platforms enhance recommendation quality by combining click‑through‑rate models with multi‑objective optimization of watch time and completion rate, using sample reweighting and weighted logistic regression to balance perceived and real relevance while improving offline AUC and online user engagement.
Meitu cut short‑video CDN bandwidth by over 50 % while preserving quality, reduced error rates from 1 % to 0.1 %, and achieved millisecond‑level startup for most videos by deploying a data‑driven fusion scheduler, multi‑CDN failover, FastDNS, buffer control, moov‑relocation and player‑tuning techniques.
The article chronicles Zhang Yiming's rise from a modest engineering student to the founder of ByteDance, detailing his four entrepreneurial ventures, the algorithmic breakthroughs behind Toutiao, the launch of Douyin, and his five concrete management principles for building a fast‑growing tech company.
Tencent's video team shares how they grew daily QQ Space video plays from 50 million to over a billion by redesigning the mobile playback architecture, implementing bandwidth and cost controls, accelerating start‑up times, adopting H.265, and building robust monitoring and error‑handling pipelines.
The talk demonstrates how Meipai, a leading Chinese short‑video community, leverages large‑scale text mining and machine‑learning techniques—ranging from anti‑spam filtering to AI‑enhanced search—to enrich captions, comments, and metadata, improve user experience, and inspire further research on text data in video platforms.
The article details the evolution, system architecture, backend technologies, high‑availability designs, data handling, upload and playback workflows, and operational monitoring of the Miaopai short‑video service, illustrating how it scales to hundreds of millions of users under heavy load.
This article outlines Meipai’s evolution into a billion‑user short‑video platform, detailing its high‑availability, scalable architecture, service discovery via etcd, data storage challenges, CDN and cloud‑storage redundancy, fault‑tolerance mechanisms, and future directions such as H.265 adoption and P2P‑CDN hybrid delivery.
