Skip to main content

Runtime Packs: FFmpeg And Python

Some workflows need more than Premiere's own capabilities: extracting frames for visual analysis, processing media files, or running data-heavy logic. Runtime Packs give Automation Agent controlled access to external tools for exactly those jobs.

A Runtime Pack is a verified, self-contained tool installation that Automation Agent manages for you — no manual FFmpeg builds, no PATH editing, no Python environment setup.

Currently available:

  • FFmpeg Pack: media analysis and processing — for example exporting specific frames from clips so a workflow can analyze them visually.
  • Python Pack: a managed Python runtime for workflows that need real data processing — for example transcript batching and validation in the library's transcript tools.

Why This Matters​

External tools extend what an automation can do well beyond timeline edits. A concrete example from the Library: Reconstruct Edit From Reference indexes selected source media, matches every shot from a rendered reference, and places reliable original source intervals back into an editable Premiere timeline. The related framing workflow can then recover Position and Scale from the approved render. No chat assistant confined to Premiere's own feature set can do that kind of media-analysis job.

Because your agent is a general-purpose agent, workflows can combine these tools with everything else: Premiere state, transcripts, files on disk, and external production context.

Installing Packs​

  1. Open the Runtime Packs tab in the Automation Agent panel.
  2. Click Install recommended tools. Automation Agent finds the right packs for your machine and asks before installing anything.
  3. Alternatively, import a signed pack archive or folder manually — useful for machines without direct internet access.

Installed packs are listed with their version and can be tested or removed from the same tab.

Runtime Packs are installed in Premiere's app-specific UXP data store. Premiere Pro and Premiere Pro Beta therefore keep separate installations, and different host profiles or major versions may do the same. If you switch to another Premiere host and a previously installed pack appears missing, open Runtime Packs in that host and install or import it there as well.

When A Workflow Needs A Pack​

Workflows declare their requirements, and Automation Agent checks them before the run (preflight):

  • If a required pack is missing, the run does not start halfway and fail — the panel reports what is missing and points you to the Runtime Packs tab.
  • MCP and non-interactive runs fail closed when a required tool is not configured, instead of guessing at system binaries.

You do not need to install packs speculatively. Install them when a workflow asks for them, or install the recommended set once and forget about it.

Trust And Verification​

External executables deserve scrutiny, so Runtime Packs are signed and verified:

  • Official packs are verified mamoworld downloads. The Runtime Packs tab shows the verification state of every installed pack.
  • Imports without valid verification data are clearly labeled as not verified.
  • Unsigned local installs are supported for development, and are labeled as such.

Pack executables run under the same execution model as everything else in Automation Agent: workflows still need the matching permissions, and filesystem access stays limited to what you allow.

Offline And Restricted Environments​

Runtime Packs are separate from AI setup: they are local tools, not AI providers. A restricted production machine can use signed pack archives imported from a file, and saved workflows that combine packs with a local LM Studio Workflow AI can run fully offline.

See High-Security and Restricted-Network Environments and Developing Offline Workflows Online.