Honorable Mention & Blog Post Award, Global AI Hackathon Series with Qwen Cloud
Anyone who has generated AI video knows the sinking feeling: the clip looks gorgeous, and it's completely wrong. The camera move you asked for never happens. The shot flickers. The duration is off, or the colors drift. You watch it ten times, regenerate, watch again. Dailies is the end of that loop — automated quality control for AI-generated video. It turns a plain-language brief into a shot pipeline, measures each take against explicit requirements, and even attempts repairs when a shot fails.
The core problem is that a convincing-looking clip can silently miss the brief, and catching that miss currently means a human reviewing every single take. For creators and marketing teams, that's repetitive, soul-crushing work that scales badly.
Dailies converts subjective "is this good?" into repeatable checks with measurable results and evidence frames. And it makes a sharp architectural choice: its deterministic checks use no model tokens at all and work on video from any generator, not just one vendor's. That means teams can catch failures before assembling a final cut, and spend their human attention on creative decisions instead of frame-by-frame policing.
A Qwen agent converts the user's brief into validated shot parameters, which the server expands into a pipeline. The user reviews the shot plan and stills before any video is generated — catching misunderstandings while they're still free to fix. Wan then generates the clips, and OpenCV checks the supported properties: duration, brightness, flicker, scene cuts, camera motion, and palette. Qwen-VL layers on advisory checks for visual meaning and continuity — the things pixel math can't judge.
When a blocking check fails, the system doesn't just flag it: Qwen revises the prompt and Wan attempts a bounded retake, which must itself pass verification before it's allowed to replace the original. The app presents results, evidence, and generation costs transparently, and caching reuses identical requests so you never pay twice for the same shot.
Qwen-plus uses function calling to author validated run parameters while the server constructs the pipeline graph. Qwen-VL supplies advisory visual assessments, Wan generates drafts and frame-anchored retakes through DashScope's asynchronous task API, and Qwen-TTS supplies narration. Deterministic OpenCV checks form the blocking verification layer at zero model-token cost — the smart economic split at the heart of the project. Content-addressed caching lets identical requests reuse generated media.
The same verification engine is exposed three ways — as a native function-calling tool, a Qwen-Agent custom tool, and an MCP server consumed by a Qwen agent — so other pipelines can borrow the quality gate. Alibaba Cloud Simple Application Server hosted the Docker Compose application (nginx, FastAPI, React, and PostgreSQL), and a private OSS bucket stored published media.
Kaung Zin Hein is a solo full-stack developer working with TypeScript, Python, and AI agents. He built Dailies — the web application, verification engine, model integrations, deployment, and accompanying technical write-up — to explore a practical question: how can teams check whether AI-generated video actually meets their requirements? His focus throughout was on evidence, and on being honest about the limits of what AI judgments can establish in generative-media workflows.
Explore the project: https://www.youtube.com/watch?v=9uQ33_IBb_c
This project was developed as part of the Global AI Hackathon Series with Qwen Cloud.
Qwen Memory Agent: Treating Memory as a Measurable Engineering Problem
144 posts | 29 followers
FollowJustin See - June 19, 2026
Community Builder - May 15, 2026
Jeff wong - November 10, 2020
Alibaba Cloud Big Data and AI - May 27, 2026
Alibaba Cloud Native Community - October 10, 2026
Alibaba Cloud Native Community - August 12, 2026
144 posts | 29 followers
Follow
Token Plan
Build more, spend less. One plan, every modality.
Learn More
QwenWork
QwenWork is dedicated to helping employees strengthen their professional competitiveness in the AI era and to enabling enterprises to improve organizational effectiveness.
Learn More
Qwen
Full-range, open-source, multimodal, and multi-functional
Learn More
Alibaba Cloud Model Studio
A one-stop generative AI platform to build intelligent applications that understand your business, based on Qwen model series such as Qwen-Max and other popular models
Learn MoreMore Posts by Alibaba Cloud Project Hub
