How to Use GPT-6 Astra for Video Editing

Set GPT-6 Astra as your model

Every workflow below starts in Daydream with Astra. Open your project in Daydream and set the model to GPT-6 Astra. Daydream makes the edit on a real timeline you can still edit manually, so Astra helps to automate the tedious work, but you keep creative control. All five workflows run on other models too, but the notes on quality below are specific to Astra.

1. Create custom motion graphics

Astra is one of the best models available today at creating motion graphics. It can model objects, animals, and abstract ideas to visualize whatever you are trying to communicate in your video. In our tests, Claude Fable 5.1 does a better job with lighting, but Astra builds more complex scenes and objects.

To try it, set your model to Astra in Daydream and describe the graphic you want: "build a motion graphic to end the video using our logo." Watch it back, then give Astra feedback on the animation, colors, and style until it looks the way you want.

Here we tested the same one-line prompt against both models. Fable 5.1 has more complex lighting, with a sunset gradient and a golden hour glow. Astra's is flatter, but it put more into the objects themselves, giving the car windows, door seams, and spoked wheels, and putting a sailboat out on the water.

2. Build a product launch video from one prompt

In our tests, Astra made the most engaging product launch videos from a single prompt. We gave it our website and a few basic product details in one message in Daydream and let it build a quick launch video from there. Most models generate these without much motion, so the result feels more like a slideshow than a video. Astra was the best at matching the energy and movement in a launch video.

If you try this, give it your brand details and assets up front: logos, colors, and fonts. That is the biggest difference between a generic result and something that looks like it came from your team.

3. Cut bad takes, silences, and filler words

Astra can also clean up bad takes and dead air. Drag your footage into the chat or import it into your project, then ask it to remove duplicate takes, silences, or filler words. It transcribes the footage, reads the transcript to find the best parts, and cuts everything else.

Afterwards, toggle on "Show cuts" in the Transcript Editor to review what it removed and make further edits by hand. The transcript cleanup guide walks through the whole flow.

4. Clean up breathing sounds with audio keyframes

If you find a lot of breathing sounds in your long interviews or voiceovers, you can also use Astra to help you clean them up. Cutting them out can impact the natural pacing, so the better fix is to lower the volume during each breath with an audio keyframe. Doing that by hand is slow, especially if you're editing a podcast or a long interview.

With Daydream, you can just ask: "clean up the breathing sounds in my voiceover." The agent will review the transcript and the audio track and add keyframes to duck the volume over the breaths. You will see the keyframes it added directly in the timeline, and they stay fully editable, so you can drag them earlier or later, or change how far the volume drops.

5. Search through your b-roll library

You could hand Astra the raw frames of your footage, but that gets expensive quickly. Daydream indexes your b-roll on import instead, working out what happens and when, so the search is already done by the time you are editing. Astra only has to retrieve the particular clip you asked for.

So you can ask for footage the way you would describe it out loud: "add the shot of the coffee being poured" or "put b-roll under this whole section." It is faster and cheaper than sending frames to the model, and your footage stays on your machine. The same trick is what makes short-form editing with AI fast.

FAQ