How to Make Lower Thirds [2026]: 10 Steps, 20 Min

Lower thirds are the name bars and caption strips that sit in the bottom part of the frame and tell viewers who is speaking, where they are or what a clip covers. Done well, lower thirds go unnoticed. Done badly, they clip at the screen edge, vanish before anyone reads them or cover a speaker’s chin. This tutorial builds one clean lower third in three editors, with exact margins and frame counts, in 10 steps per editor and roughly 20 minutes for a first build.

Versions matter because every editor moves its menus. This guide follows Adobe Premiere 26.5.2 (the September 2026 release), DaVinci Resolve 21.1.1, which Blackmagic’s own readme documents, and the current CapCut desktop app. Resolve has a free edition, Premiere costs $22.99 a month on Adobe’s annual plan billed monthly, and CapCut has a free tier. Pricing and menu labels below are dated October 30, 2026.

Lower thirds belong to the wider family of on-screen text, so if you also build title cards, credits or caption styles, the lower thirds and titles section collects those guides. Here the focus stays on one job: a name-and-title graphic that stays readable on a phone, sits inside the safe area and is timed to the frame.

What Lower Thirds Do in a Video

A lower third is on-screen text or graphics that identify a speaker, a location, an organization or a topic. The most common version is two lines: a name on top and a role underneath. Guides aimed at YouTubers, teachers and business video teams all describe the same core use, which is introducing a speaker without stopping the footage to explain who they are.

The name comes from placement. The graphic occupies the lower part of the frame, below the eyeline and above the very bottom edge, so the face and the action stay visible. That placement is a design choice, not a law. In a vertical clip the lower zone moves up to clear the app interface, and in a talking-head video you may flip it to the top if the speaker holds a product in the bottom half of the shot.

Lower thirds differ from two things people often mix them up with. Captions transcribe speech, word by word, and run continuously. A title card or title sequence fills the screen and sets the mood, like the smoke-style opener in the epic titles sequence. A lower third does neither. It supplies one short fact, shows it for a few seconds and leaves.

Five variants cover most real projects:

• Name and role. Two lines, shown once per speaker, 3 to 6 seconds.
• Location or date. One line, often in news-style b-roll, tied to a cut.
• Topic or chapter label. One line that changes at each section of a tutorial.
• Handle or call to action. A social handle or channel name, shown sparingly.
• Source credit. A short line for archive footage, a photo or a quoted clip.

The rest of this article treats the first variant as the build target. Once you have a name-and-role graphic with correct margins and timing, the other four are a matter of deleting a line or changing a label. You will also generate a cue sheet with a short script, so every graphic in the video follows the same timing rule instead of whatever felt right at 1 a.m.

Prerequisites and Versions (October 2026)

You need footage with at least one speaker, one editor from the table below and a text editor for the cue script. Python 3.8 or newer runs the script, since it uses f-strings and the standard library only. ffmpeg is optional and used for guide overlays and export checks.

Adobe’s product pages now call the desktop app Adobe Premiere, while its release documentation still says Premiere desktop and its developer pages say Premiere Pro. This guide writes Premiere for short. The newest release listed in Adobe’s notes is 26.5.2 from September 2026, and Adobe advertises a seven-day free trial if you want to follow along without a plan.

One caution on Resolve. A 21.1.2 maintenance build was reported on October 27, 2026, though the official readme this article checked covers 21.1.1. Open Help > About in your copy and use whichever 21.1.x you have, because the lower third workflow below does not depend on the patch number.

CapCut announced new AI tools and a ChatGPT plugin at CapCut World in Los Angeles on October 27, 2026. Nothing in this tutorial needs them. Menu names in CapCut shift between desktop, web and mobile, so the CapCut section describes actions first and labels second.

Safe Areas for Lower Thirds

Keep every pixel of text inside a margin you set yourself, and confirm it against the editor’s own safe-area overlay. The idea behind safe areas is old and still holds: content near the frame edge can be cropped or hidden on some screens, so text belongs inside an inner region. What has no single answer is the exact percentage. Each editor, sequence setting and platform draws the guide differently, and no current official source backs the claim that one universal number applies everywhere.

So this article uses working margins instead of a quoted standard. They are the author’s house rules, chosen so they are easy to remember and easy to scale: 5 percent on the sides for landscape video, and the bottom edge of the graphic at least 10 percent above the bottom of the frame. Treat them as a starting point and widen them if your platform puts interface elements near the edges.

Landscape 16:9

At 1920×1080, a 5 percent side margin is 96 pixels, so the left edge of the bar and text starts at x=96. The bottom edge of the graphic sits at y=972, which leaves 108 pixels below it. That puts the whole graphic low in the frame, clear of faces in a standard medium shot.

Left alignment works better than centering for name bars. Eyes read left to right, and a left-aligned bar reads as a label for the person in the shot. If the speaker sits on the left third of the frame, mirror the layout: right-align the bar so it does not cover their shoulder.

Vertical 9:16 and 4:5

Short-form apps draw their own captions, usernames and buttons over the bottom of the frame. For 1080×1920 the house rule is a 10 percent side margin (108 pixels) and a graphic that ends about 25 percent above the bottom edge, at y=1440. For 1080×1350 (4:5), end it 20 percent up, at y=1080. Test with a real upload to a private or unlisted post before you batch-export a whole series, because each app changes its interface over time.

The 4K row is simply the 1080p row doubled. Build at the canvas you deliver in and scale the numbers. If you build a 1080p graphic and drop it into a 4K timeline, check that your editor scales vector shapes cleanly instead of enlarging a bitmap.

Do not rely on the editor’s viewer alone. A safe-area overlay in a program monitor tells you where the guide sits in that editor, not how a TV, a laptop or a phone app will crop the frame. Export a short test, play it on a phone and a monitor, and confirm no letter touches an edge. The project section later in this article includes a phone-size proxy command for exactly that check.

Lower Thirds Timing: In, Hold and Out

Every lower third has three phases. It animates in, holds still while viewers read, then animates out. Keep the in and out moves short and spend the rest of the time on the hold. A long move starts to feel slow, and a hold that is too short leaves slow readers behind.

Use frames for the animation and seconds for the hold. Pick one move length in frames and keep it the same across every graphic in the video. Mixed move lengths look like a mistake even when viewers cannot say why.

For the hold, base the time on the number of words on screen and give slower readers a little extra. It is an editorial rule of thumb, not a published standard. It errs on the side of a slightly long hold, because a graphic that lingers costs less than one that disappears mid-read.

Placement in the timeline matters as much as length. Cue the graphic one to three seconds after the speaker first appears, never in the first second of a shot. Viewers need a moment to see the face before they look for a label. Show it once per speaker, and repeat only if the same person returns after a long gap.

Also avoid starting a lower third on a cut. Landing the animation 6 to 12 frames after the cut lets the new shot register first. And never let one graphic overlap another. If two speakers appear back to back, end the first one before the second begins, or the screen turns into a pile of boxes.

Design Rules That Survive a Phone Screen

Most lower thirds fail on small screens, not big ones. The text is too thin, the bar is too transparent or the title line is so small that it turns into a smear after compression. Fix those three problems and the rest of the design matters far less.

Use one typeface family and two weights. A bold weight for the name and a regular weight for the role gives hierarchy without a second font. Sans-serif faces with open letterforms survive compression better than thin serifs or script fonts. Keep the name at 48 pixels or larger on a 1080p canvas, and do not let the role line drop below 32.

Contrast comes from the bar, not the footage behind it. Adjust the opacity, color and background treatment whenever the text blends into the picture. A solid or near-solid bar (85 percent opacity in the spec used here) is more reliable than a faint tint, because footage changes under the graphic while the bar stays fixed. White text on a dark bar is the safest default. If your brand color is light, put the text in dark gray instead of forcing white on pastel.

Keep animation in service of reading. A plain slide or wipe from the left reads as confident, while bounces, spins and glows look dated fast. Font size and animation speed are the two adjustments that most change how clearly a lower third lands, so change those before you touch the color.

Finally, write the text for the screen. Titles like Senior Vice President of Global Marketing Operations overflow any bar. Shorten to Head of Marketing, and let the speaker say the rest. A name and a role should fit in a bar no wider than 900 pixels at 1080p. If the role needs 12 or more words, it belongs in the video description.

Set Up the Project: Spec, Guides and Cue Sheet

Before opening an editor, write the design down. A short spec file keeps every graphic in the video identical, and it doubles as a handoff note if someone else edits the series. This one matches the 1080p row of the safe-area table.

{
  "canvas": { "width": 1920, "height": 1080, "fps": "30000/1001" },
  "margins_px": { "left": 96, "right": 96, "bottom": 108 },
  "text": { "name_px": 48, "title_px": 32, "name_weight": "bold", "title_weight": "regular" },
  "bar": { "height_px": 120, "max_width_px": 900, "opacity": 0.85 },
  "timing": { "anim_s": 0.4, "hold_min_s": 3.0, "hold_max_s": 6.0 }
}

Next, read the facts about your footage. A timeline that does not match the source frame rate is the quiet cause of half the timing complaints later on. This ffprobe command prints the resolution, frame rate and pixel format of the clip.

ffprobe -v error -select_streams v:0 \
  -show_entries stream=width,height,r_frame_rate,pix_fmt \
  -of default=noprint_wrappers=1 interview.mp4

# example output
width=1920
height=1080
pix_fmt=yuv420p
r_frame_rate=30000/1001

Then build a transparent guide image. The command below draws guide rectangles on an empty RGBA frame: the 5 percent margin and the lower third zone. Adjust the numbers if you work at a different canvas size. Editors without a built-in safe-area overlay, CapCut included, can use this PNG on the top video track.

ffmpeg -f lavfi -i "color=c=black@0:s=1920x1080:r=1,format=rgba" \
  -vf "drawbox=x=96:y=54:w=1728:h=972:[email protected]:t=4:replace=1,\
drawbox=x=192:y=108:w=1536:h=864:[email protected]:t=4:replace=1,\
drawbox=x=96:y=852:w=900:h=120:[email protected]:t=2:replace=1" \
  -frames:v 1 guide_1920x1080.png

You can also burn the guide into a preview copy of your footage to check framing before you build anything. The overlay filter stacks the PNG on every frame, and the audio is copied untouched.

ffmpeg -i interview.mp4 -i guide_1920x1080.png \
  -filter_complex "overlay=0:0" \
  -c:v libx264 -crf 20 -c:a copy guide_preview.mp4

Last, list your speakers. Create a CSV with a name, a role and the second where each graphic should start. The cue script reads it, applies the hold formula and prints frame-accurate cues.

name,title,start
Dana Reyes,Lead Producer,4.0
Marcus Webb,Head of Post-Production and Color,12.5
Priya Nair,Founder,14.0
Tom Okafor,Sound Designer at Northline Audio Studio,40.0

Here is the script. It uses only the Python standard library. It computes the hold from the word count, converts seconds to frames at the frame rate you pass in, flags overlaps and graphics that run past the end of the video, and writes both a CSV and an SRT file.

#!/usr/bin/env python3
import argparse
import csv
from fractions import Fraction


def parse_args():
    p = argparse.ArgumentParser(description="Build a lower third cue sheet")
    p.add_argument("csv_in", help="CSV with columns: name,title,start (seconds)")
    p.add_argument("--fps", default="30000/1001", help="for example 30, 25 or 30000/1001")
    p.add_argument("--duration", type=float, default=None, help="video length in seconds")
    p.add_argument("--anim", type=float, default=0.4, help="in and out animation, seconds")
    p.add_argument("--out", default="cues", help="output file prefix")
    return p.parse_args()


def timecode(frame, nominal):
    sec, ff = divmod(frame, nominal)
    mm, ss = divmod(sec, 60)
    hh, mm = divmod(mm, 60)
    return f"{hh:02d}:{mm:02d}:{ss:02d}:{ff:02d}"


def srt_time(seconds):
    ms = round(seconds * 1000)
    hh, ms = divmod(ms, 3600000)
    mm, ms = divmod(ms, 60000)
    ss, ms = divmod(ms, 1000)
    return f"{hh:02d}:{mm:02d}:{ss:02d},{ms:03d}"


def hold_seconds(name, title):
    words = len((name + " " + title).split())
    return min(6.0, max(3.0, 1.5 + 0.35 * words))


def main():
    args = parse_args()
    fps = float(Fraction(args.fps))
    nominal = round(fps)
    with open(args.csv_in, newline="", encoding="utf-8") as fh:
        rows = sorted(csv.DictReader(fh), key=lambda r: float(r["start"]))

    cues = []
    prev_end = 0.0
    for n, r in enumerate(rows, start=1):
        start = float(r["start"])
        hold = hold_seconds(r["name"], r["title"])
        end = start + args.anim + hold + args.anim
        f_in = round(start * fps)
        f_hold = f_in + round(args.anim * fps)
        f_out = f_hold + round(hold * fps)
        f_end = f_out + round(args.anim * fps)
        notes = []
        if start < prev_end:
            notes.append("OVERLAP")
        if args.duration is not None and end > args.duration:
            notes.append("PAST END")
        prev_end = end
        cues.append((n, r["name"], r["title"], start, end, f_in, f_hold, f_out, f_end, " ".join(notes)))

    with open(args.out + ".csv", "w", newline="", encoding="utf-8") as fh:
        w = csv.writer(fh)
        w.writerow(["cue", "name", "title", "in_tc", "hold_tc", "out_tc", "end_tc", "frames", "notes"])
        for c in cues:
            w.writerow([c[0], c[1], c[2], timecode(c[5], nominal), timecode(c[6], nominal),
                        timecode(c[7], nominal), timecode(c[8], nominal), c[8] - c[5], c[9]])

    with open(args.out + ".srt", "w", encoding="utf-8") as fh:
        for c in cues:
            fh.write(f"{c[0]}\n{srt_time(c[3])} --> {srt_time(c[4])}\n{c[1]} / {c[2]}\n\n")

    for c in cues:
        print(f"cue {c[0]}: {timecode(c[5], nominal)} to {timecode(c[8], nominal)}  {c[8] - c[5]:3d} frames  {c[1]} {c[9]}")


if __name__ == "__main__":
    main()

Run it against the sample file. The sample includes two deliberate mistakes so you can see the warnings: cue 3 starts while cue 2 is still on screen, and cue 4 runs past the end of the video.

python3 cues.py speakers.csv --fps 30000/1001 --duration 45 --out cues

cue 1: 00:00:04:00 to 00:00:07:24  114 frames  Dana Reyes 
cue 2: 00:00:12:15 to 00:00:17:07  142 frames  Marcus Webb 
cue 3: 00:00:14:00 to 00:00:17:24  114 frames  Priya Nair OVERLAP
cue 4: 00:00:39:29 to 00:00:45:02  153 frames  Tom Okafor PAST END

The CSV it writes gives you in, hold, out and end timecodes for each cue. Check them against your editor’s own timecode display, especially on a drop-frame timeline. When in doubt, trust the frame counts, or use the SRT file, which works in seconds.

cue,name,title,in_tc,hold_tc,out_tc,end_tc,frames,notes
1,Dana Reyes,Lead Producer,00:00:04:00,00:00:04:12,00:00:07:12,00:00:07:24,114,
2,Marcus Webb,Head of Post-Production and Color,00:00:12:15,00:00:12:27,00:00:16:25,00:00:17:07,142,
3,Priya Nair,Founder,00:00:14:00,00:00:14:12,00:00:17:12,00:00:17:24,114,OVERLAP
4,Tom Okafor,Sound Designer at Northline Audio Studio,00:00:39:29,00:00:40:11,00:00:44:20,00:00:45:02,153,PAST END

Fix the sample before moving on. Push cue 3 out to second 18 and pull cue 4 back to second 38, then rerun the script until no warnings appear. In a real video, use the SRT as a reference track: drop it on the timeline, and the cue text shows you where each graphic belongs.

Frame by Frame: Lower Thirds in Premiere

Premiere builds lower thirds in the Essential Graphics panel, opened from Window > Essential Graphics. You can place a built-in or downloaded .mogrt template on a video track above the footage and then change its text, font, colors and motion settings. Building your own takes the steps below, and the result exports as a template you can reuse. Adobe moves labels between builds, so use the panel search if a menu name differs.

1. Match the sequence to the footage. Right-click the clip in the Project panel and choose New Sequence From Clip. Confirm the frame size and rate in Sequence > Sequence Settings against your ffprobe output.
2. Turn on the safe-area guide. In the Program Monitor, open the settings (wrench) menu and enable Safe Margins. Drop guide_1920x1080.png on V3 as a second reference, and disable it before export.
3. Create the name text. Select the Type tool (T), click in the Program Monitor near the lower left, and type the first name. The text lands as a graphic clip on V2 and appears in the Essential Graphics Edit tab.
4. Add the role line. Click again and type the second line as a separate text layer. Set the name to 48 px bold and the role to 32 px regular, per the spec file.
5. Draw the bar. Pick the Rectangle tool, drag a shape behind the text, and set its fill to your brand color at 85 percent opacity. Size it to 120 px tall. Drag the shape layer below the text layers in the layer list.
6. Pin the bar to the text. Select the shape, open Responsive Design – Position, and pin it to the name layer with padding on all sides. Now a long name widens the bar instead of overflowing it.
7. Position by the guide. Set the group’s left edge at x=96 and the bottom at y=972. Check that nothing crosses the red 5 percent line.
8. Animate in and out. In Effect Controls, keyframe Position (or Opacity) at frame 0, frame 12, then again 12 frames before the clip ends and at the last frame. Right-click the keyframes and apply ease in and ease out so the motion decelerates into the hold.
9. Protect the hold. Use Responsive Design – Time to mark the intro and outro segments. Premiere then stretches only the middle when you change the clip’s duration, so the moves stay at 12 frames.
10. Export and reuse. Right-click the graphic and choose Export As Motion Graphics Template. Expose the name and role layers in the Edit tab so editors can change them without opening the layer stack. Save, then drag the template on V2 above the next speaker.

To place the template on a later speaker, drop it on a track above the footage, then change the wording from the Essential Graphics Edit tab. Drag the clip edge to the cue length from your sheet. Play the sequence at full resolution to check the timing, because a lower third that looks right frozen can still feel late in motion.

Three details save time on a series. Name the template by size (for example LT_1080p_NameRole) so nobody puts a 1080p file in a 4K sequence. Keep one template per aspect ratio rather than relying on Responsive Design to rescale the graphic. And store templates in a shared folder if a team edits the same channel, so everyone uses the same version.

Frame by Frame: Lower Thirds in DaVinci Resolve

Resolve gives you two routes. The quick one is a built-in title template from the Effects Library. The flexible one is a custom graphic in Fusion, built from a Text+ node and a few nodes for the bar. The release notes for 21.1.1 reportedly mention improvements involving Text+, MultiText, transitions, Fusion effects and OGraf animation, but the steps below work the same across 21.x. The steps assume the Edit and Fusion pages, and the free edition covers all of it.

1. Set the timeline first. Open File > Project Settings > Master Settings and set the timeline resolution and frame rate to match the footage, before you add clips. Changing the rate after the timeline exists is awkward.
2. Turn on the safe area. Enable the Safe Area overlay from the viewer or the View menu. Put the guide PNG on V3 if you prefer your own margins.
3. Try a built-in template. In the Effects Library, open Titles and look for the lower third entries (names vary by version). Drag one onto V2, above the footage, and edit the text in the Inspector.
4. Start a custom build. Drag a Text+ title onto V2, select it, and open the Fusion page. You will see a Text+ node connected to MediaOut.
5. Add the bar. Add a Background node, then attach a Rectangle mask to it. Set the rectangle’s width and height to the size of your bar, converted into Fusion’s 0 to 1 coordinates, then adjust by eye against the guide, and move it to the lower left.
6. Set the text. In the Text+ node, type the name, choose the bold weight, and set the size so cap height matches the 48 px spec. Duplicate the node for the role line at 32 px, and position both inside the bar using the guide.
7. Merge in order. Add a Merge node with the bar as background and the name text as foreground, then a second Merge for the role text. The stack order is bar first, then text.
8. Add a Transform and animate it. Place a Transform node after the last Merge and connect it to MediaOut. Right-click its Center input, choose Expression, and paste the slide-in expression shown below.
9. Return to the Edit page. Trim the title clip to the cue length. Duplicate it for each speaker and edit the two Text+ nodes in each copy’s Fusion page, or turn the node group into a macro (see the advanced tips).
10. Render a check. Park the playhead 15 frames into the clip, confirm the bar is in position and not clipped, then export from the Deliver page.

Fusion expressions use Lua, and time is the current frame. The expression below slides the whole graphic in from the left over 12 frames, easing to a stop. The offset starts at a fraction of the frame width and falls to zero with easing.

-- Transform1 > Center (right-click > Expression, then paste)
-- Slides the whole lower third in from the left over 12 frames, easing out
Point(0.5 - (1 - math.min(time / 12, 1)) ^ 3 * 0.4, 0.5)

To animate out as well, extend the expression so a second offset grows during the final 12 frames. It uses the comp’s render end as the reference, so it follows the clip length when you trim it. If Fusion flags an error, fall back to manual keyframes at the same frames and the result will match.

-- Same slide-in, plus a 12-frame slide-out that starts 12 frames before the end
Point(0.5 - (1 - math.min(time / 12, 1)) ^ 3 * 0.4 - (math.max(0, time - (comp.RenderEnd - 12)) / 12) ^ 3 * 0.4, 0.5)

A note on time. In a Fusion title the comp’s frame counter starts at the first frame of the clip, so time equals 0 at the head of the graphic. If you nest the graphic inside a compound clip that starts later, subtract the start frame from time in the expression.

Resolve’s free edition limits output to Ultra HD or lower, which is no restriction for lower thirds in 1080p or 4K delivery. Studio adds higher resolutions and multi-GPU processing, which matter for other jobs. Studio costs $295 as a one-time purchase, with a temporary $236 price at Blackmagic resellers through November 20, 2026.

Frame by Frame: Lower Thirds in CapCut

CapCut builds a lower third from a text layer, a shape and an animation preset, and it works with no special template. The steps below describe the desktop editor. Labels differ on web and mobile, so look for the action first. CapCut’s public pages did not confirm a built-in safe-area overlay, which is why step 2 uses the guide PNG you made earlier.

1. Create the project and set the ratio. Pick 16:9 for landscape or 9:16 for vertical before you place anything. Changing the ratio later rescales every layer.
2. Add the guide. Import guide_1920x1080.png (or the 1080×1920 version) and place it on the top track for the full clip length. You will delete it before export.
3. Add the name text. Open the Text tab, add default text to the timeline above the footage, and type the name. Pick a clear sans-serif font and a bold style.
4. Add the role text. Add a second text layer for the role. Set it to a regular weight and about two-thirds the size of the name.
5. Add the bar. From the shapes or elements area, drop in a rectangle, send it behind the text, and set the fill color and an opacity near 85 percent.
6. Align to the guide. Use the position fields in the right-hand panel to place the bar so its left edge sits on the red margin line and its bottom on the cyan zone. Nudge the text inside with the same fields.
7. Group the layers. Select the three layers and, if your build offers it, create a compound clip so they move together. If not, trim all three to the same length.
8. Add animation. Select the bar and open Animation. Choose a slide or wipe as the In animation and a matching Out animation. Keep both durations short, well under half a second. Add a gentle text fade as the text’s In if you want a softer finish.
9. Time it from the cue sheet. Trim the group to the cue length, long enough for a viewer to read a two-line graphic comfortably. Copy and paste it for each speaker and edit the text.
10. Hide the guide and export. Delete or mute the guide track, export at 1080p with the frame rate matching the source, and check one frame against the guide before you publish.

CapCut’s pricing is the least fixed of the three. CapCut describes Free and paid Pro tiers, with a reference price near $19.99 a month in some markets and local pricing elsewhere, so check the number inside your own account. According to CapCut’s own pages, Pro includes 4K/60 export, no-watermark export and exclusive templates, and the upgraded Pro plan lists 1,200 AI credits and 1 TB of cloud storage. A lower third needs none of the AI credits.

Treat templates with care. CapCut’s Text tab and template library offer ready-made lower thirds, and they are fast, but many are built for a different safe margin and a different aspect ratio than your canvas. Check any template against your guide PNG. If text touches the red line, move or shrink it before using it.

Render Settings and Export Checks

Export at the resolution and frame rate of the timeline, not at whatever the dialog suggests. A mismatch forces a scale or a frame rate conversion that softens thin text and adds judder to animated moves. The table lists the settings used in this tutorial.

From the Vipid archive: the original intro themes list their resolution, length and frame rate, which is a useful model for documenting your own lower third pack.

Check the render before you upload it. Pull a frame from the middle of the first lower third, confirm no text touches a margin, and scale the video down to phone size to judge legibility. These two commands do both. If you cannot run ffmpeg, scrub your exported file in a media player and check a paused frame at full size.

# grab the frame where cue 1 should be fully in (about 0.5 s after it starts)
ffmpeg -ss 00:00:04.5 -i final.mp4 -frames:v 1 check_cue1.png

# phone-size proxy, no audio, to judge legibility at arm's length
ffmpeg -i final.mp4 -vf "scale=640:-2" -c:v libx264 -crf 30 -an phone_test.mp4

Transparent clips deserve a short word. Exporting the lower third alone with an alpha channel lets you reuse it in any editor that reads ProRes 4444, which includes the three covered here on most setups. Support differs by edition and platform, so test one clip in the target editor before you export a whole pack. Resolve’s codec list in the Deliver page shows what your edition can write.

Common Pitfalls With Lower Thirds

These six mistakes show up in nearly every first attempt. Each one has a quick fix, and most come from skipping the setup stage rather than from any editor limitation.

• Text too close to the edge. Content near the frame edge can be cut off on some screens. Keep the 5 percent margin, and check it against a guide on top of the footage, not by eye.
• A hold shorter than reading time. Three-word names get 3 seconds, and longer lines get more. A graphic that clears at the one-second mark reads as a flash.
• Slow animation. Moves above half a second drag. Keep them well under that and use easing, not a linear move.
• Weak contrast. White text over a pale shirt disappears. Put a solid bar behind the text and raise its opacity before you change the font.
• Inconsistent styling across the video. Different fonts, margins or timing between graphics make the edit look improvised. Use the spec file and one template for the whole video.
• Frame rate mismatch. Set the timeline to the same frame rate as your footage, since a mismatch can make animated moves look uneven. Read the rate with ffprobe and set the timeline to match before you start.

One more problem is quieter: covering the speaker. A bar that sits across the chin or a hand-held product undermines the shot. If the subject is low in the frame, move the graphic to the top, or shift it to the other side. A good lower third gets out of the picture’s way.

Also resist the urge to decorate. Drop shadows, outlines, glows and gradients add noise and make compression artifacts worse. A flat bar, one accent line and two lines of text cover almost every interview and tutorial format.

Troubleshooting Lower Thirds

When a lower third misbehaves, the cause is usually a track order, a missing font or a mismatched setting. Work through these in order before rebuilding anything.

1. The graphic does not show. It sits on a track below the footage. Move it to a higher track, such as V2 above V1. In Resolve, also confirm the Fusion comp’s last node connects to MediaOut.
2. A Premiere template shows a missing font warning. The .mogrt uses a font that is not installed. Install the font, or choose a replacement in the Essential Graphics panel and recheck the bar padding, since a different font changes the text width.
3. A template will not load. Confirm the file ends in .mogrt and sits in the folder your editor reads. Restart the app after copying files in. Templates made in a newer build may not open in an older one, so update before you blame the file.
4. The text overflows the bar. The text is not pinned to the shape. Use Responsive Design – Position in Premiere, shorten the line, or widen the shape in Fusion or CapCut.
5. The animation stutters in playback. Render the section or use proxies, and move project and media to a local drive. Adobe documents network and removable media issues that cause dropped frames in Premiere.
6. The Fusion expression turns red. A typo, a missing parenthesis or a wrong input type causes this. The Center input expects a Point, so the expression must return Point(x, y). Retype it on one line, or fall back to keyframes.
7. Exported text looks softer than the preview. The export scaled the picture, or the bar sits on a half pixel. Match the export size to the timeline and move shapes by whole pixels.
8. Gradient bars show banding after export. Compressed 8-bit video struggles with smooth gradients. Use a flat fill, or add a very light grain over the gradient.
9. The cue times drift by a few frames. The cue script writes non-drop timecode. On a drop-frame timeline the display differs by a few frames per minute. Use the frame counts or the SRT file.
10. The lower third sits under app captions on a phone. Lift the graphic into the lower-middle zone for vertical video, or move it to the top. Test with a private upload before publishing.
11. The CapCut export differs from the preview. Check that the canvas ratio matches the export ratio, and confirm the export resolution. Free and paid tiers can differ in export options, so check your plan’s limits.

Advanced Tips for Better Lower Thirds

Once the basic build works, a few habits turn it into a reusable system. None of them require extra software, and each saves minutes per video that add up over a season of episodes.

Publish controls in Resolve. Select the nodes in the Fusion page, create a macro, and publish the two text inputs. The macro then shows editable name and role fields in the Edit page Inspector, so you stop opening Fusion for every speaker. Save it as a title template for the next project.

Build a pack, not a single graphic. Make four variants from one master: name and role, one-line location, chapter label and handle. Keep margins, fonts and timing identical. A pack looks like a brand, while four unrelated graphics look like four experiments.

Match the brand once. Pick one accent color and one neutral for the bar. Set the accent for a thin line on the left edge and let the bar stay dark. That single line signals the brand without costing contrast.

Add a second language line. For subtitled interviews, a third text row can carry a translation of the role. Test the longest case first. German and Finnish roles run long, and the bar must grow without hitting the 5 percent margin.

Use markers for review. Drop timeline markers at every cue from the CSV, or import the SRT as a reference track. Reviewers can step marker to marker and check each graphic against the margin guide instead of scrubbing blindly.

Automate the boring part. The cue script is the start of a pipeline. Extend it to emit a JSON file for each cue, and a template script in your editor of choice can read it. Even without scripting inside the editor, a clean CSV removes the guesswork about which name goes where.

Mind accessibility. Viewers with low vision or small screens benefit from larger text and strong contrast. If your channel gets a lot of mobile traffic, raise the minimum text size by 10 to 20 percent over the spec above and test again.

The Complete Working Project

Put the pieces together in one folder. The structure below keeps footage, guides, cues, specs, graphics and exports apart, so a template update never touches a finished video. It is the layout the earlier commands assumed.

lower-thirds-project/
  footage/interview.mp4
  guides/guide_1920x1080.png
  cues/speakers.csv
  cues/cues.py
  cues/cues.csv        (generated)
  cues/cues.srt        (generated)
  spec/lower-third.json
  graphics/lower-third.mogrt    (Premiere export)
  exports/final.mp4

The run order, start to finish, takes about 20 minutes for the first video and far less for later ones:

1. Run ffprobe on the footage and copy the width, height and frame rate into the spec file.
2. Generate the guide PNG with ffmpeg at the same size.
3. Fill speakers.csv with names, roles and rough start times.
4. Run cues.py with your frame rate and video length, and fix any OVERLAP or PAST END flags.
5. Build the graphic once in your editor, following the 10 steps for it above.
6. Place or paste the graphic at each cue timecode and edit the text.
7. Hide the guide, export, then run the proof-frame and phone-size commands.
8. Fix whatever the proof shows, re-export, and archive the template with the spec file.

Expected result: each speaker’s lower third slides in over 12 frames, holds for 90 to 180 frames depending on length, and slides out over 12 frames, with all text inside the 96-pixel side margin and the bar ending at y=972. If your proof frame shows that, the graphic is ready to ship.

Keep the spec file and cue script with the project and reuse them on the next episode. The first run teaches you the settings. After that, a new video is a new CSV, a rerun of the script and a paste of the template at each cue.

Before You Render

What is a lower third?

A lower third is an on-screen text graphic in the bottom part of the frame that identifies a speaker, location, organization or topic. The usual form is a name over a role, shown for a few seconds and animated in and out.

How long should a lower third stay on screen?

Give a short name and role a brief hold, and longer text a bit more, with a quick move in and out. Preview in the sequence, since too short leaves readers behind and too long distracts from the video.

Where do lower thirds go to stay inside the safe area?

Keep text inside the editor’s title-safe region and inside your own margin. This guide uses 5 percent on the sides and a bottom edge at least 10 percent above the frame bottom for landscape, with more room at the bottom for vertical video. Exact guide percentages differ by editor and platform.

Can I make lower thirds free?

Yes. DaVinci Resolve has a free edition that handles everything in this tutorial, and CapCut has a free tier. Premiere is a paid app with a seven-day free trial, according to Adobe’s current pages.

Should I export lower thirds as transparent clips?

Only if you want to reuse one graphic across editors. A transparent ProRes 4444 clip keeps the animation fixed, but editable text in a template or macro is usually more practical for name changes. Test the clip in the target editor first.

Why do my lower thirds look blurry after export?

Usually the export scaled the frame, the bar sits on half-pixel coordinates or compression is too heavy for thin text. Match the export size to the timeline, snap shapes to whole pixels and use a bolder weight for the role line.

Do lower thirds work the same in vertical video?

The structure is the same, but the position changes. Vertical platforms overlay captions and buttons near the bottom, so lift the graphic about 25 percent above the bottom edge and test it with a real upload before you publish a series.