How to add b-roll to a 2–3 hour YouTube video in Premiere Pro
Summarize withChatGPTClaudePerplexity
A two-hour documentary, history video or sleep story has a simple problem: the voiceover is done, but the screen needs something to show for every second of it. If the shot changes about every six seconds, a two-hour video needs around 1,200 clips. Found, downloaded, trimmed and placed one by one, that is days of work before you even start on music and titles.
This guide shows a workflow for b-roll on ultra long-form videos in Premiere Pro, first by hand and then the automatic way, and the mistakes that only show up in long videos.
1. Set up the timeline
Put the voiceover on its own audio track, for example A2, and lock it so nothing you do to the b-roll can move or cut it. Keep the b-roll on V1 with its sound on A1, and mute A1: stock clips often have wind, music or traffic that fights with your voice. If you talk on camera, keep your facecam on V1 and put the b-roll on V2 above it, only where a sentence has something to show.
Use a sequence that matches your final export (for example 1920×1080 at 29.97 or 25 frames per second) and decide it before you place anything. Changing the frame rate of a long sequence later shifts hundreds of cuts.
2. Work sentence by sentence
A cut that lands on the start of a sentence feels natural; a cut in the middle of a word feels like a mistake. Transcribe the voiceover first (Premiere Pro's Speech to Text does it inside Premiere), then use the sentence starts as your cut points. One clip per sentence is a good default. Very short sentences can share a clip, and long ones can be split at a comma.
For slow, calm videos such as sleep stories or relaxing documentaries, fewer and longer clips work better: let a clip play in full and only change it when the sentence changes.
3. Search for what can be filmed
Stock sites can only show things a camera can film. “The universe is about 13.8 billion years old” has no clip, but galaxy, night sky and stars do. For each sentence, pick the one or two concrete things in it and search for those. When a sentence is completely abstract, fall back to footage that fits the topic of the whole video, so the viewer is never left with a black screen.
Leave out clips with text, logos or watermarks, and clips in the wrong orientation. A vertical phone clip in a landscape video stands out at once.
4. Don't repeat clips too often
In a ten-minute video nobody notices a repeat. In a two-hour video the same galaxy shot five times in a row is the first thing comments mention. A simple rule keeps it fresh: use a clip at most three times, at least ten minutes apart, and start it from a different moment each time. Keeping a list of the clips you used makes this possible.
5. Stay in sync for hours
Long videos expose tiny errors. If every clip ends a frame late, the b-roll is several seconds behind the voice after an hour. Cut each clip to an exact number of frames, place it on the frame where its sentence starts, and check the end of the timeline against the end of the voiceover. Conform clips to your sequence's frame rate and size before placing them, so Premiere does not have to guess.
Download each clip in the smallest size that still covers your sequence: a 4K file for a 1080p video only fills your drive. Over a thousand clips, that difference is many gigabytes.
Work in sections of 15 to 30 minutes and save after each one.
6. Keep the credits as you go
Pixabay and Pexels do not require credit, but crediting creators is good practice, and a list of where every clip came from helps if a video is ever claimed. Note the creator, the site and the link for each clip while you place it; writing it afterwards for a thousand clips is hopeless. See our guide to free stock footage licences and credits.
The automatic way
VoxVision is a free Premiere Pro plugin that does steps 2 to 6 for you, and it was made for exactly these long videos:
- It transcribes the voiceover with Premiere Pro's Speech to Text, or with a speech model on your PC. The audio never leaves your computer.
- It picks search words for every sentence and searches Pixabay (and Pexels, if you have a key) with your own free API key, then filters out clips with text, logos and the wrong orientation before anything is downloaded.
- It downloads only the clips it uses, cuts them to the exact frame and places them on their own tracks in batches, so the first shots appear while the rest is still being found. You can stop and continue at any time.
- Each clip is used at most three times, at least ten minutes apart, from a different moment of the clip. You can change these rules.
- Your voiceover, your facecam and clips you placed yourself are never moved, cut or muted. Clips from your own footage folder are used first.
- You can replace or block any clip, and you get a credits list ready for the video description.
To try it: download VoxVision (Windows, Premiere Pro 2026), sign in for free, add your free Pixabay API key, put your voiceover on the timeline and press Start auto edit. More about how it handles 2–3 hour videos.
Questions editors ask
How many b-roll clips does a 2-hour video need?
With a new shot about every six seconds, around 1,200. Slow videos with long, whole clips need fewer.
How long should a b-roll clip be?
As long as its sentence, often 3 to 10 seconds. Very short sentences can share a clip and long ones can be split at a comma.
How do I stop the b-roll drifting out of sync in a long video?
Cut every clip to an exact number of frames, place it on the frame where its sentence starts, and keep the sequence frame rate fixed from the start.
Can Premiere Pro add b-roll automatically?
Not by itself. A plugin such as VoxVision reads the voiceover and places a matching stock clip under every sentence on its own tracks.