Vchitect
Shusheng DreamBuilder (Vchitect) is a video generation model launched by the Shanghai AI Laboratory. The official statement says that Vchitect 2.0 supports generation of videos from 5 to 20 seconds, flexible aspect ratios, and super-resolution frame interpolation, and also has the capability of generating videos from images.
Updated: 2026-10-26
What is Vchitect?
Shusheng·DreamBuilder (Vchitect) is a video generation model launched by the Shanghai AI Laboratory. The official website positions Vchitect 2.0 as an advanced generative model for video creation. It focuses on long-duration generation and flexible aspect ratios. According to the official statement, it integrates super-resolution and frame interpolation and allows users to adjust the degree of content correction. In addition to the model itself, the VBench evaluation suite open-sourced by the team has also been adopted by multiple mainstream video models and has become a common evaluation reference in this field.
Who is it for?
- Video creation teams and independent creators: Need to quickly produce longer clips starting from text or images to validate creative directions.
- Vertical short video operators: Require flexible aspect ratio output to avoid repeated cropping and reprocessing to adapt to different platforms.
- AI video researchers and engineering teams: They need a reference open-source evaluation system to conduct horizontal comparisons of the performance of different video generation models.
- Pre-production creative teams in film and advertising: Use generated clips for concept demonstrations and storyboard rehearsals, reducing communication costs before on-location shooting.
Key features
6Long-duration video generation
The official states that Vchitect 2.0 can generate videos ranging from 5 to 20 seconds in duration, which is more suitable for presenting complete shots or short narrative segments compared to common clips of just a few seconds.
Flexible Aspect Ratio
The official states that it supports generating videos with any aspect ratio, making it convenient to adapt to different delivery formats such as vertical short videos and horizontal long videos.
Super-resolution and Frame Interpolation
The official states that the model integrates super-resolution and frame interpolation capabilities to enhance the resolution of generated frames and the smoothness of motion.
Content Modification Control
The official states that users can adjust the degree of content modification to strike a balance between preserving creativity and aligning closely with the prompt.
Image-to-Video
The official claims that it can convert static photos or design drafts into 5 to 10-second videos, bringing existing image materials to life.
VBench Evaluation Suite
The official open-sourced an automatic video generation evaluation suite and supports one-click evaluation deployment. The official stated that it has been adopted by models such as Gen-3, Kling, and OpenSora.
How to use Vchitect
- Before generation, determine the aspect ratio and duration of the target platform to reduce image quality loss caused by subsequent cropping and frame interpolation.
- When it comes to open-source models and evaluation suites, the specific licenses and usage scope shall be subject to the official repository and page declarations, and each item must be confirmed before commercial use.
- There is still uncertainty regarding the copyright and commercial boundaries of generated content.
- It is recommended to confirm the license terms and consult professionals before external release and commercial deployment.
- AI-generated videos may still have distortions in human movements and text.
- manual screening and post-editing should be retained before official delivery.
- The mainstream working method for AI video generation is still "short clips + post-editing splicing": the duration of a single generated clip is limited.
- For complex narratives, it is recommended to split them into multiple clips based on storyboards for generation, and then unify the rhythm and soundtrack with video editing software.
Traffic & popularity
Data month Sep 2026
Vchitect receives about 2.4M visits per month.
Monthly visits
12 monthsEstimated from third-party traffic data.
Similar tools
Tencent Hunyuan AI VideoTencent Hunyuan has launched a text-to-video capability. Officially, its open-source video models rank among the top in terms of parameter scale and performance. It supports text-to-video generation, native shot cutting and continuous actions, and also provides voice-driven digital human images and automatic dubbing capabilities.
SkyReelsSkyReels is an AI video generation platform launched by Kunlun Tech. Centered on storytelling narrative, it provides capabilities such as text-to-video, image-to-video, and multimodal reference-to-video. According to the official statement, it can generate both video and audio simultaneously, and supports character consistency and plot continuity across storyboards.
GenmoGenmo is a research-driven open-source video generation lab. With the Mochi text-to-video model and open-world models, it enables users to create high-quality videos locally or in the playground.
SparkVidAn online creation platform integrating multiple mainstream text-to-video models. Its official website states that it can call engines such as Seedance, Kling, and Veo on the same interface to generate videos from text or images, and supports reference video to motion conversion, camera control, and natural language video editing.
MedeoMedeo is a tool for creating AI videos through conversation. It can generate finished videos from text, audio, images, videos or links, and is built with functions such as clipping, outfit changing, watermark removal and quality enhancement.
Pollo AIAn AI creative suite that aggregates multiple mainstream video and image models. It enables the creation of text-to-video, image-to-video, digital human, audio, and marketing materials all within a single workspace, supports up to 4K output, and is ideal for creators who do not wish to be tied to a single model ecosystem.