🪄 Alpha functionality: This functionality is still being developed and is expected to change over time. Right now we only enable it for a limited number of customers. If you're interested, reach out to your Customer Success Manager for more details and eligibility requirements.

--

Welcome to Video Ad Assembly!

Video Ad Assembly allows users to go from raw clips or existing long-form videos to short-form videos with AI.

• How it works:

  • User inputs starter clips (existing long-form videos, unedited clips, etc.)

  • AI analyzes the video (visual, text, audio, etc.) and breaks it into clips

  • User prompts AI for what kind and what length of new variation to create

  • AI outputs new variants, 3 at a time. User can prompt to request more up to 15 total

  • User can further edit scripts and output to Template Editor

• This feature will benefit anyone who wants to save time in video storyline creation, especially if they want to

  • Cut long form videos down to short form

  • Make iterations of existing short-form videos (that has enough content to make sensible re-cuts)

  • Turn a large set of raw clips into a sensibly scripted video

Accessing Video Ad Assembly

You can access video ad assembly from two places:

1. AI Playground: feature card
   
2. Media Library: select New -> Video Template -> AI Ad Assembly 

Note that when you start Video Ad Assembly, you'll be asked to select an aspect ratio. This can be changed later easily, so don't worry about it too much yet.

Subscription and Non-Subscription customers will have different levels of access.

Plan your input videos and goals

Before you actually start the flow within the UI, we recommend that you plan which videos you want to input, and the type of result that you get out.

Keep in mind that AI has control over only two types of iteration on your video: cutting up and rearranging clips, or adding new text overlays. It will not generate new clips for you, and it cannot do anything that a human cannot do. So if your input video is a perfectly-cut video that you cannot imagine changing in any way, you may be disappointed in the outcome.

Ideal candidates for inputs and outputs are as follows:

1. Long-form videos -> Short-form videos. Videos that are 30s+ or even 60s+ long, with many  potential iterations that you could imagine producing as a human. This tends to work really well with videos with talking audio, such as talking head UGC content or commercials with a talking-led story.
   1. A way to hack this if you don't have long videos: upload multiple similar short-form videos, so AI can mix clips across those videos.
2.  Raw clips -> assembled short-form video. Raw clips, ideally of similar types (e.g. all talking videos, or all nature/app/product shots). Note that if the videos don't have existing talking audio, you will need to add text overlays to guide the story.

Also keep in mind you have a 5 minute input limit, so choose what you want to upload.

Ready? Let's dive into the flow.

Upload and analyze videos

Select one or more videos, up to 5 mins in video length in total. Press "analyze" when you are ready to move forward.

Analysis takes about 2x the length of the input video, so make sure your video selections are final before you move forward.

Review Scene Splits, Re-Split as Needed

After analysis, we will automatically split your video into clips. The clips will be split based on a mix of visuals (visual transitions) and talking audio transcription (we re-combine clips to avoid sentences being cut off mid-transition). Review the clips by clicking on and watching them.

Not happy with the cuts? Try re-splitting using the optional settings.

• Try transcription-only method if your video has primarily talking audio, and you want to cut basd on this
• Try visuals-only method if you video doesn't have much talking audio, and you want to cut based on visual transitions
• Try changing minimum clip length if you're finding that the clips are too long or short. The possible range here is 1-10s.

You can also individually ignore clips with the "eye" icon.

Ready to move forward?

Rescript Videos

Now Video Ad Assembly has all the raw ingredients: clips, analyzed and ready to rearrange.

Tell Video Ad Assembly what you're looking for in the video brief. Example prompts:

• make iterations
• make iterations that advertise [your brand here]
• make iterations that focus on the value proposition of adventure
• make variants, one that focuses on adventure, one that focuses on luxury, one that focuses on a mix
• add a storyline that advertises [your brand here] to families

You can also set a min and max video length. Note these are approximate -- AI will not always follow these perfectly, especially if you're trying to prompt it to output <15s. The sweet spot tends to be 15-30s requests.

Finally, consider adding "suggest text overlays". If this is turned on, AI will suggest (and later add) text overlays to the video. We do NOT recommend this for videos with existing talking audios, but we DO recommend this for videos without existing talking audio or otherwise without an existing strong story.

Press "generate scripts". After a short wait, you'll see 3 scripts appear. You can edit these scripts by dragging different clips in and out! Want to revert your changes? Press the "refresh" button to rest back to the original AI-created script.

You can also re-prompt and keep generating, up to 15 scripts. 

Choose scripts to output and optional settings

Ready? Choose which scripts to output! All are included by default, but you can uncheck "keep this script in export" to ignore it.

Also consider these optional settings:

• On each video, you can select "remove background music, only keep talking audio". This will allow you to do what it sounds like -- remove background music! This helps disparate clips flow better together -- you can always add music back later in editor. Remember this is a setting on each individual video, not a universal check.
  
  
• For all videos, select the output aspect ratio. You can force one aspect ratio for all, or choose "match source" to have the aspect ratio set by the first clip within each output video.
  

Press "turn videos into templates", and boom! You now have videos in Template Editor that you can further improve.

If you get an error with videos rendering correctly, just save and refresh the page.