video-use Default Subtitle Settings: Chunking, Case, and Placement

By default, video-use creates 2-word uppercase subtitle chunks that break on punctuation, render in Helvetica 18 Bold with 90px bottom margin, and are overlaid last in the ffmpeg pipeline.

This article breaks down the hard-coded subtitle defaults in the browser-use/video-use repository, explaining how automatic caption generation works when you use the --build-subtitles flag. All settings are defined in helpers/render.py and require no configuration to produce platform-ready captions.

Default Subtitle Chunking Behavior

The chunking algorithm in video-use prioritizes readability and timing precision. In helpers/render.py (lines 44-71), the build_master_srt() function implements a two-word maximum with early termination on punctuation.

Chunk Size and Punctuation Breaks

The default chunk size is 2 words per subtitle cue. However, if a word ends with any punctuation mark in PUNCT_BREAK, the chunk terminates immediately:

PUNCT_BREAK = (".", ",", "!", "?", ";", ":")

# Chunking logic from helpers/render.py

chunks = []
current = []
for w in words_in_seg:
    current.append(w)
    if len(current) >= 2 or w["text"].endswith(tuple(PUNCT_BREAK)):
        chunks.append(current)
        current = []

This means a phrase like "Wait, what?" splits into two cues: ["Wait,"] and ["what?"] — not one four-word cue.

Text Case Conversion

All subtitle text is converted to uppercase using Python's str.upper():

text = " ".join(w["text"] for w in chunk).upper()

This happens during SRT generation, before any styling is applied. The uppercase default ensures maximum legibility on mobile screens and social feeds.

Subtitle Placement and Visual Style

The SUB_FORCE_STYLE string in helpers/render.py (lines 51-56) hard-codes the placement and appearance:

SUB_FORCE_STYLE = (
    "FontName=Helvetica,FontSize=18,Bold=1,"
    "PrimaryColour=&H00FFFFFF,OutlineColour=&H00000000,BackColour=&H00000000,"
    "BorderStyle=1,Outline=2,Shadow=0,"
    "Alignment=2,MarginV=90"
)

Property	Default Value	Purpose
Font	Helvetica 18 Bold	Clean, platform-neutral typeface
Color	White (&H00FFFFFF)	Maximum contrast on varied backgrounds
Outline	2px black	Subtle definition without heavy shadow
Alignment	2	Bottom-center positioning
MarginV	90px	Safe zone above TikTok/IG Reels/YouTube Shorts UI chrome

The 90px vertical margin is specifically calibrated for vertical video. At typical 9:16 resolutions, this places captions approximately 30% up from the bottom — clear of profile pictures, like buttons, and comment prompts.

Composite Order: Subtitles on Top

In helpers/render.py (lines 38-44), the ffmpeg filter graph appends subtitles last:

filter_parts.append(
    f"{current}subtitles='{subs_abs}':force_style='{SUB_FORCE_STYLE}'[outv]"
)

This guarantees that captions render above all video overlays, color grades, and animated elements. The subtitles filter receives the fully composited video stream, ensuring text remains crisp and unobscured.

Using the Default Subtitle Pipeline

To trigger automatic subtitle generation with these defaults:

from helpers.render import build_master_srt

# This creates master.srt with 2-word uppercase chunks

build_master_srt(edl, edit_dir, edit_dir / "master.srt")

Or via command line:

python video_use.py --build-subtitles input.mp4

No additional parameters are required. The function reads the transcript from the EDL (edit decision list), applies the chunking rules, and writes a standard SRT file that the render pipeline consumes.

Customization Notes

These defaults are hard-coded constants in helpers/render.py. To modify behavior, you must edit the source:

• Change 2 in len(current) >= 2 to adjust word count
• Remove .upper() to preserve original case
• Edit SUB_FORCE_STYLE for different fonts, colors, or positioning
• Adjust MarginV for different aspect ratios or platform requirements

The repository does not expose configuration flags for these values. This design choice ensures consistent output across all renders in a production workflow.

Summary

• Chunking: 2-word maximum, early break on .,!?;: punctuation
• Case: Uppercase via str.upper() in build_master_srt()
• Placement: Bottom-center with 90px margin, rendered last in ffmpeg pipeline
• Style: Helvetica 18 Bold, white with 2px black outline
• Source: All defaults defined in helpers/render.py

Frequently Asked Questions

How do I disable automatic chunking and use full sentences?

You must modify helpers/render.py. Remove or increase the len(current) >= 2 condition and the PUNCT_BREAK check. The current implementation has no runtime flag for sentence-level chunking.

Why is the default margin set to 90 pixels?

The 90px MarginV value positions subtitles in the safe zone for vertical video platforms. According to the video-use source code, this clears interface elements on TikTok, Instagram Reels, and YouTube Shorts without manual adjustment per platform.

Can I change the font without editing the source code?

No. The SUB_FORCE_STYLE string is a module-level constant in helpers/render.py. You must edit this file to change font family, size, color, or positioning. There is no command-line interface for subtitle styling.

Does the uppercase conversion affect non-English languages?

The str.upper() method uses Python's Unicode-aware case mapping. For most Latin and Cyrillic scripts, this works as expected. Complex scripts (Arabic, Devanagari) may not have uppercase forms, in which case the text passes through unchanged.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}