With the release of version 25.3, Adobe has introduced a new AI-powered Search feature in Premiere Pro. In addition to searching through text and metadata, the Search panel can now identify visual content in images and video clips, for example, footage of a lake surrounded by hills in the early morning. Imported video and audio files can also be automatically transcribed in the background, making them searchable through the same Search panel.
How does it work?
The visual search feature is powered by Adobe’s new media intelligence analysis. When an asset is imported into a project, it is automatically analysed using on-device AI models. Once the analysis is complete, users can easily locate assets in the new Search panel by entering relevant objects or phrases. Adobe recommends using multi-word, conversational descriptions, similar to how you would speak to a person, including details like lighting or camera angle.
The results of this AI analysis are stored either in the users’ media cache folder or as a sidecar file next to the asset. If saved in the media cache, the analysis data will be accessible across any project that uses the asset, but only on the computer where the analysis was performed. If the results of the analysis are saved as a sidecar file next to the asset, they will be available in any project that the asset is imported into and on any computer.
How can this feature impact performance?
After updating Premiere Pro to version 25.3, the media intelligence feature is enabled by default. This means that all clips already in the project, as well as any newly imported clips, will be automatically analysed. In a shared storage environment, this can lead to increased network traffic, as each clip must be pulled from the central storage to the local machine for analysis.
By default, the analysis results are saved in the user’s media cache location. Since other users do not have access to this metadata, each user will independently analyse the same files. This redundancy can generate significant and unnecessary network traffic, especially in larger collaborative environments. We have seen more than 700 MB/s of traffic from a single 10 Gbit client while analysing footage.
The solution
The new AI-powered Search is a powerful addition to Premiere Pro, offering editors a faster way to locate specific footage. However, if users aren’t exactly aware of how it operates, it can lead to performance issues in shared storage environments. To avoid these problems, we recommend one of the following two approaches:
Disable media intelligence analysis (if not needed)
If you’re already using another method to locate footage, such as the Search function in the Media Library Panel, you may not need the new AI Search. In that case, you can disable it to prevent unnecessary processing and network traffic.
To do this, go to Preferences > Media Analysis & Transcription in Premiere Pro, and uncheck both “Analyze all imported media to visually search your projects” and “Automatically transcribe files.”
Note: This setting must be disabled on each individual workstation.
Use AI search efficiently (if you want to keep it enabled)
If you want to take advantage of the new AI Search, it’s important to set it up efficiently for a shared environment. Ensure that the analysis results are saved Next to the media in a sidecar file. This allows metadata to be shared across different systems.
Optionally, you can designate a single computer on the network to handle the analysis. This way, redundant processing is avoided, and only one workstation performs the AI tasks, reducing both resource usage and network load.
Check out some of our other blogs about Premiere Pro:
Adobe Premiere Pro – Performance and Troubleshooting Guide
Using Premiere Pro Productions with ELEMENTS Storage
Comparing the Auto Reframe Functions in DaVinci Resolve and Premiere Pro
Everything You Need to Know About the Proxy Workflow in Adobe Premiere Pro
Exploring the New Scene Cut Detection Features of DaVinci Resolve and Adobe Premiere Pro
