How to Use Focused Mode with Start and End Timestamps in Claude Video

The watch skill enters focused mode when you supply --start and/or --end flags, automatically increasing the frame extraction density for that specific segment while filtering transcripts and timestamps to match the original video timeline.

The claude-video repository by bradautomates provides a powerful watch skill that analyzes video content using Claude's computer vision capabilities. When you need to examine a specific moment rather than process an entire video, focused mode with start and end timestamps lets you target precise segments with optimized frame sampling. This mode automatically applies a denser frame budget and constrains all outputs to your specified time window.

How Focused Mode Works

When you provide either --start or --end parameters, the script evaluates focused = start_sec is not None or end_sec is not None in skills/watch/scripts/watch.py (lines 55‑58). This boolean triggers a specialized extraction pipeline that prioritizes detail within your defined boundaries.

Parsing Timestamps with parse_time

User-provided time strings are converted to seconds by the parse_time function defined in skills/watch/scripts/frames.py (lines 55‑71). This utility handles multiple human-readable formats including HH:MM:SS, MM:SS, or raw integer seconds.

Calculating the Effective Range

In skills/watch/scripts/watch.py (lines 40‑53), the script computes effective_start, effective_end, and effective_duration based on your provided values. If you omit --end, the skill analyzes from your start point to the video's conclusion. If you omit --start, processing begins at 00:00.

Optimizing Frame Density with auto_fps_focus

The primary advantage of focused mode lies in the auto_fps_focus implementation in skills/watch/scripts/frames.py (lines 41‑58). Unlike the standard auto_fps function, this specialized routine allocates a denser frame budget for short clips while strictly respecting the global 2 FPS ceiling. This ensures maximum visual detail from your targeted segment without exceeding processing limits.

Extracting Frames and Filtering Content

The extraction functions in skills/watch/scripts/frames.py (lines 86‑92) receive start_seconds and end_seconds arguments, passing them directly to ffmpeg to process only the requested slice. Simultaneously, skills/watch/scripts/watch.py (lines 63‑66) invokes filter_range to trim any available transcript segments to the same time window, ensuring textual analysis aligns perfectly with the extracted visual frames.

Output Reporting

The final report includes a "Focus range" line and preserves timestamps relative to the original video timeline rather than offsetting from the clip start, as implemented in skills/watch/scripts/watch.py (lines 78‑84).

Usage Examples

Here are practical commands demonstrating focused mode with start and end timestamps:

# Analyze the last 10 seconds of a 1-minute video

python3 "$SKILL_DIR/scripts/watch.py" video.mp4 --start 50 --end 60

# Zoom into a 30-second window at maximum 2 FPS

python3 "$SKILL_DIR/scripts/watch.py" "$URL" --start 2:15 --end 2:45 --fps 2

# Extract from 1 hour 12 minutes to the end of the video

python3 "$SKILL_DIR/scripts/watch.py" "$URL" --start 1:12:00

Summary

• Focused mode activates automatically when you supply --start or --end flags to the watch skill.
• The parse_time function in skills/watch/scripts/frames.py converts human-readable timestamps to seconds.
• auto_fps_focus allocates a denser frame budget for short segments while respecting the 2 FPS global limit.
• ffmpeg receives time boundaries directly via start_seconds and end_seconds, and transcripts are filtered via filter_range to match the visual range.
• Output preserves original video timestamps and displays the focused range for verification.

Frequently Asked Questions

What timestamp formats does focused mode support?

The parse_time function accepts multiple formats including raw seconds (e.g., 50), minutes:seconds (e.g., 2:15), and hours:minutes:seconds (e.g., 1:12:00). All values are normalized to seconds before the effective_start and effective_end calculations begin.

Can I use focused mode without specifying both start and end times?

Yes. If you provide only --start, the skill analyzes from that point to the video's end. If you provide only --end, it processes from the beginning (00:00) up to that timestamp. The logic in skills/watch/scripts/watch.py (lines 40‑53) handles these partial ranges by deriving effective_duration from the video metadata.

Does focused mode affect transcript analysis?

Yes. When a transcript is available, the script trims it to match your specified time window using filter_range (lines 63‑66 in skills/watch/scripts/watch.py). This ensures that any text-based analysis corresponds exactly to the visual frames being extracted and analyzed.

What is the maximum frame rate in focused mode?

Even in focused mode, the skill respects the global 2 FPS ceiling defined by the selected detail mode. The auto_fps_focus function in skills/watch/scripts/frames.py optimizes the frame distribution within this hard limit, selecting a higher effective rate for short clips without ever exceeding 2 frames per second.

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