Make Claude Edit Your Talking Videos

Edit my talking video. The file is in this folder: [your video file name, like IMG_1234.MOV]

I want 2 things done to it:
1. Cut out the dead pauses so it moves faster.
2. Add captions of what I say, burned into the video.

Do it this way:

1. Check first. See if this computer has ffmpeg (with its subtitles filter), Python 3.9 or newer, and auto-editor. Tell me in plain words what's missing, what you'd install, and about how big the downloads are. Then stop and wait for my yes before you install anything.

2. Keep it all in this folder, so nothing else on my computer changes. Programs go in a "tools" folder here. Python packages go in a virtual environment here called .venv. Don't use sudo or anything that needs my password.
- ffmpeg: if it's missing, use Homebrew if I already have it (Mac) or winget (Windows). If not, download a ready-made build from a site listed on ffmpeg.org/download.html into the tools folder. If none of those fits my computer, use another well-known build and tell me which site you used.
- auto-editor: download the official build for my computer from github.com/WyattBlue/auto-editor/releases into the tools folder. It's not on pip anymore.
- Captions: install faster-whisper into the .venv with pip. Use the "small" model on the CPU with compute_type "int8", and keep the model download in the tools folder too. If faster-whisper fails with an error about "metadata_errors", install av version 18 or lower in the .venv and try again.
- If Python is missing, tell me the simplest way to get it and wait.

3. Never change my original video. Save everything new in a folder called "edited".

4. Cut the pauses with auto-editor. Use a margin of 0.2 seconds so no word gets clipped. Save it as edited/(video name)-cut.mp4.

5. Make captions from the cut video with faster-whisper, using word timestamps. Make short caption lines of 2 to 4 words, each one on screen while I say it. Save them as edited/(video name)-captions.srt.

6. Burn the captions into a finished video: edited/(video name)-final.mp4. Use H.264 video and AAC sound so it plays everywhere. Make it vertical, 1080 wide by 1920 tall. If my video is sideways (wider than tall), ask me before you crop it. Phone videos are often HDR. If mine is, convert it to normal colors first, or the captions may not show and the colors can look washed out.
Caption look: [bold white words with a black outline]. Big enough to read on a phone, centered, in the lower third, but above the bottom 15% of the screen where app buttons sit. One line at a time.

7. Check your work before you tell me it's done:
- Pull 3 still frames from the finished video that have captions on them, and look at them. The captions must be big, easy to read, and fully on screen. If not, fix it and redo the video.
- Compare the length before and after, and count the cuts.
- Tell me if any quiet stretch over 1 second is still in, and where.
- Read the captions back. List any words you're not sure about (names, brand names, words the speech tool wasn't sure of) so I can check them.

8. Save the steps as a script in this folder called edit_video.py, so next time I can just say "run edit_video.py on my new video".

When you're done, tell me in plain words: where the files are, how long the video was before and after, how many pauses you cut, and which caption words I should double check.

Change this in my edited video: [what to fix, like "there's still a long pause about 17 seconds in, cut it" or "the cuts feel rushed, leave a bit more room around my words" or "in the captions, Jon should be John"]

Change only that. Save a new finished video and captions file with -v2 at the end of the names, and keep the old ones. Then check 3 frames again and tell me what changed. If the fix should stick for future videos, update edit_video.py too, but keep its normal file names.

Run edit_video.py on [your new video file name]. Then check 3 frames and tell me which caption words I should double check.