Upload footage to a video project

Add original video files that the assistant can inspect, select from, and edit within the correct listing project.

Before you start

Use footage you own or are authorized to use. Confirm that it shows the same property as the project's linked listing. Avoid exposing residents, documents, access codes, or other private information.

Keep the original files until you have verified the cloud copy and the final download.

Choose the right upload route

Project Footage upload is for general source videos:

• MP4, MOV/QuickTime, or WebM.
• Up to 64 MiB per file.
• Up to 180 seconds per source file.
• Maximum source dimension: 3840 pixels on either side.

Upload this scene from gallery is for a take attached to a shooting-script scene:

• The same supported video containers.
• Up to 64 MiB.
• Up to 65 seconds per uploaded scene take.

Record with teleprompter captures a take in the browser:

• Up to 60 seconds per recording.
• Kept video takes upload and link back to the project.

These are different limits. A two-minute clip may fit the project uploader but not the scene-gallery uploader.

Upload directly to the project

1. Open the correct project and check the property in the header.
2. Select Footage.
3. Select Choose a video to upload.
4. Choose one file and wait for processing to finish.
5. Look for Video added to the project and a playable clip in Project footage.
6. Play the clip to confirm the picture, orientation, and content.
7. Repeat for the remaining files.

The service checks the actual video bytes. Renaming an unsupported file to “.mp4” will not make it valid.

Upload against a script scene

1. Open the project's Scripts tab and the current script.
2. Go to On-site filming.
3. Select the scene that the clip belongs to.
4. Select Upload this scene from gallery.
5. Open Takes and verify Added to video project.
6. Select View project footage to check the clip in the original project.

This route preserves the script, version, and scene context. A preferred take helps organize the handoff; it does not replace evidence-based selection in the final edit.

Copy into ChatGPT after uploading

@bsquare-agentos Read the footage in video project “[project name or ID].” Confirm which uploaded clips are linked to it and inspect their actual video evidence.

Summarize the useful coverage, any obvious gaps, and any footage that needs a closer motion check. Use the original project assets, not listing photos, to describe what was filmed. Do not save an edit plan or render yet.

The assistant should read the uploaded project assets directly. You should not have to copy opaque asset IDs into a normal conversation.

If uploading through ChatGPT is supported by your connection, the file must still become an owned AgentOS video asset and be linked to the project. A chat attachment or arbitrary video URL alone is not a linked project asset.

If an upload fails

The file is too large or long: Keep the original, then export a smaller or shorter copy in a supported video format. Preserve useful resolution and avoid repeated compression.

The upload result is unconfirmed: Check the project's footage list before uploading again. If it is missing, retry the same file in the same project.

The take is saved in the cloud but absent from the project: In Takes, use Retry adding to video project.

The file plays on your phone but is rejected: Its container, duration, frame-rate metadata, rotation, or dimensions may not pass server validation. Export a standard MP4 copy or use another supported source.

There is only a device copy: Download a backup before clearing browser storage or changing devices.

Check the result

You are ready to edit when the intended clips appear and play in Footage. Uploading alone does not trim, arrange, stabilize, or render them.

Current Director output is vertical 9:16. Landscape footage can be cropped, but the assistant must check that the main subject remains visible. Original audio is currently muted by the Director edit.

Next: Create and review an AI edit.