Reels ad safe zone at 1080x1920 in pixels and the caption anchor

Meta's Instagram Reels ad specification says to leave roughly 14% of the top, 35% of the bottom and 6% of each side free of text and logos. On a 1080x1920 frame that is about 269 px at the top, 672 px at the bottom and 65 px on each side, leaving a 950 px wide by about 979 px tall box. A burned-in caption line centred at anchor_ratio 0.55 sits inside that box. The numbers below are my arithmetic on Meta's percentages.

What Meta says

The percentages come from Meta's Instagram Reels ad guide, read 2026-10-04: leave roughly 14% of the top, 35% of the bottom and 6% on each side free, so the profile icon, call to action and other interface elements do not cover key creative. The page lists the image spec as 9:16 at 1440x2560 with a 30 MB cap; the safe-zone sentence applies to any asset you place in a Reels ad. The percentages are relative, so they hold at 1080x1920 as well as at 1440x2560.

Turning that into an anchor_ratio

Video captions takes design.placement.anchor_ratio, defined in the docs as the line centre as a fraction of frame height. The renderer's own bounds are 0.05 to 0.95. Your line centre must satisfy: centre minus half the line block height is at least 269 px, and centre plus half is at most 1247 px. For a two-line block about 150 px tall, that range is a centre between 344 and 1172 px, or 0.18 to 0.61. Pick 0.55 for a margin: centre 1056 px, block from 981 to 1131 px.

Width has a sibling knob, design.typography.safe_width_ratio: 950 / 1080 is 0.88. Treat that as a starting value and test it, because the docs list the field without a range and an out-of-range number is a 400 at request time. design is not supported on the punch or tiktok-green styles.

def safe_box(w=1080, h=1920):
    top, bottom, side = round(h * 0.14), round(h * 0.35), round(w * 0.06)
    return {"top": top, "bottom": bottom, "side": side,
            "box_w": w - 2 * side, "box_h": h - top - bottom}

def anchor_range(block_px, w=1080, h=1920):
    z = safe_box(w, h)
    lo = (z["top"] + block_px / 2) / h
    hi = (h - z["bottom"] - block_px / 2) / h
    return round(lo, 2), round(hi, 2)

print(safe_box())
print(anchor_range(150))  # (0.18, 0.61)

Verify on the real file

Do not trust arithmetic alone. Burn the caption, then pull a still with video inspect at an explicit timestamp (frames: { at: [2.5] }, max_edge up to 2160) and look at it with the 35% band masked. Use seek: precise, the default, so the timestamp matches. If the geometry is wrong, restyle with source_caption_id; the docs say that re-burn reuses the word timings, with no second speech-to-text, though a restyle is still a billed render.

Meta's page covers its own placements. TikTok and YouTube publish their own safe-zone templates; do not reuse these numbers there.