Workflow guide
B-roll that proves the story instead of decorating it
Random pretty footage over narration feels like a slideshow; purposeful B-roll feels like evidence. The difference is matching each shot to a specific claim and cutting it exactly where the words need proof. This guide covers selection, timing, and finishing for coverage that earns attention.
From the Mellius product team · Updated September 28, 2026
Conspiracy theorists, you may hate this
AI Director
›
Just type what you need
- ›Remove all silent parts
- ›Add bold kinetic captions
- ›Only have Bill Ackman speaking
- ›Cut to highlights only
- ›Add cinematic transitions
- ›Match cuts to the beat
The short answer
List each claim needing visual proof, source close-ups and action shots over wide scenery, lay B-roll to cover jump cuts and illustrate nouns as spoken, keep shots two to five seconds, then match exposure and color to the main camera.
Every shot has a job
Each clip visualizes a specific word, claim, or emotion in the narration.
Cuts land on language
B-roll enters on the noun or verb it illustrates, not seconds later.
Coverage matches
Exposure, color, and sharpness feel like one shoot, not a stock patchwork.
Choose shots that prove, not fill
Watch the edit with the picture off and mark every claim a skeptic would question — statistics, processes, results, places. Those marks are your B-roll list. Favor close-ups of hands, tools, screens, and faces over generic skyline pans; specificity reads as truth while generic beauty reads as filler. Mix shot scales deliberately: a wide establishes, a medium follows action, a close-up delivers texture.
Budget realistically: plan one B-roll shot every 5–8 seconds for retention-style edits, fewer for calm explainers. Log selects with action-based labels like hands-typing-closeup rather than clip numbers so you can find them mid-edit. In Mellius you can ask for B-roll over specific claims and review the suggested coverage as editable timeline clips.
- One shot per claim a skeptic would question.
- Close-ups and action beat generic wides.
- Plan a new visual every 5–8 seconds for fast formats.
- Label selects by action, not by filename.
Cut, layer, and blend the coverage
Cut B-roll in on the word it illustrates — the frame should change as the noun lands, with audio leading by 2–5 frames (a mini J-cut) so the transition feels motivated. Keep most shots 2–5 seconds; longer only when the action itself is interesting. Layer B-roll on a track above the interview so the A-cam audio runs uninterrupted and jump cuts hide underneath the coverage.
Finish by matching: align exposure with the waveform, nudge white balance toward the A-cam, and add a touch of the project grain or sharpening so everything shares a texture. Duck location audio under narration to −20 dB or lower so it adds atmosphere without competing. The classic mistake is stacking five effects on B-roll while the interview stays flat — match first, stylize second.
- Enter on the noun; let audio lead picture by a few frames.
- Hold 2–5 seconds per shot in most pacing styles.
- Keep interview audio continuous under the coverage track.
- Match exposure and color before adding any stylization.
Compare the workflows
| Source | Best fit | Trade-off |
|---|---|---|
| Self-shot coverage | Best match and most authentic proof | Takes planning and shoot time |
| Stock libraries | Fast gap-filling at volume | Risks generic or mismatched looks |
| Assisted or AI selects | Finding coverage for long scripts | Needs human review for accuracy |
A practical sequence
Put it into practice
- 01
Map claims to shots
Mark every line needing proof and list the close-up or action that shows it.
- 02
Cut on language
Lay coverage above the A-cam, entering on nouns with short audio-led transitions.
- 03
Match and mix
Align exposure and color to the main camera and tuck location audio under narration.
Frequently asked questions
How often should B-roll appear?
For retention-paced YouTube and social video, a new visual every 5–8 seconds is a solid baseline. Slower explainers and documentaries can hold shots longer when the action earns it. Let claim density decide: every assertion that needs proof deserves coverage, regardless of the clock.
How do I hide jump cuts with B-roll?
Lay the B-roll on a track above the interview so the audio plays continuously while the picture changes. Cut away just before the jump and return just after, keeping the coverage 2–5 seconds. Matching color between the two angles makes the return feel seamless.
How do I make stock footage match my video?
Match exposure first with the waveform, then white balance toward your A-cam, then apply your project look across both. Add consistent grain and keep sharpness similar — over-sharpened stock over soft interview footage is a giveaway. When possible, pick stock with similar lighting direction to your shoot.
Continue reading
Primary sources
Product capabilities change. These official references support the comparisons above.
Describe the edit. Keep the timeline.
Mellius turns conversational direction into cuts, captions, graphics, motion, color, and sound you can still inspect and change.
Start editing free