mirror of
https://github.com/langgenius/dify-official-plugins.git
synced 2026-07-21 17:45:23 -04:00
Qwen-VL-Plus model cannot recognize video anomalies, but works fine with images #574
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @qiaofenlin on GitHub (Aug 21, 2025).
Self Checks
Dify version
1.7.2
Cloud or Self Hosted
Self Hosted (Docker)
Steps to reproduce
Use the qwen-vl-plus model from Tongyi Qianwen in Dify.
Try to recognize anomalies in a video input.
The model does not return results for video.
When using an image input instead, the model works normally.
测试-chatflow.yml
✔️ Expected Behavior
The qwen-vl-plus model should be able to recognize anomalies in videos, similar to how it successfully recognizes anomalies in images.
❌ Actual Behavior
The qwen-vl-plus model fails to recognize video anomalies
@dosubot[bot] commented on GitHub (Aug 21, 2025):
The Qwen-VL-Plus model in Dify only supports image and text input—video input and video anomaly recognition are not currently supported. This limitation is due to both the model provider's API and Dify's integration, which only handle images for multimodal tasks. Inputs outside this spec, like video files, are ignored or rejected by validation or at the provider level, and there is no explicit video error handling in the codebase.
If you need to analyze video anomalies, a practical workaround is to extract frames from your video and process those frames as images through Qwen-VL-Plus. This approach lets you leverage the model's image anomaly detection capabilities for video content. There are no documented workarounds or recent updates that add direct video support for Qwen-VL-Plus in Dify at this time [source].
To reply, just mention @dosu.
How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other
@dosubot[bot] commented on GitHub (Aug 21, 2025):
The Qwen-VL-Plus model in Dify is set up to accept video input, but it only supports videos provided as URLs—not as base64 data. If you try to send base64 video, you'll get an error asking you to set
MULTIMODAL_SEND_FORMATtourlsource.Even with the correct format, whether the model can actually recognize anomalies in videos depends on the capabilities of the Qwen-VL-Plus API/service itself—not on Dify. There’s no extra video anomaly detection logic or post-processing in Dify; it just forwards your video URL to the model source. If video anomaly detection isn’t working, it’s likely a limitation of the model or its backend service.
If you need video anomaly detection and Qwen-VL-Plus isn’t delivering, you might consider extracting frames from your video and running image anomaly detection on those frames, or using a different provider/model that explicitly supports video anomaly analysis.
Make sure your video input is a URL and not base64, and check the model’s documentation or service for details on its video analysis capabilities.
To reply, just mention @dosu.
How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other
