Configuration

Most users won't need this page — vidpipe init handles setup automatically. This is the full reference for when you want fine-grained control.

The Simple Version

Most users just need a .env file with one key:

env

OPENAI_API_KEY=sk-your-key-here

Everything else has sensible defaults. Run vidpipe init to generate a complete .env interactively.

TIP

Configuration is resolved in priority order: CLI flags → environment variables → .env file → defaults.

CLI Parameters

Positional Argument

Argument	Description
`[video-path]`	Path to a video file to process. Implies `--once` mode.

Options

Flag	Description	Default
`--watch-dir <path>`	Folder to watch for new `.mp4` recordings	`$WATCH_FOLDER` or `./watch`
`--output-dir <path>`	Base directory for processed output	`$OUTPUT_DIR` or `./recordings`
`--openai-key <key>`	OpenAI API key (Whisper + agents)	`$OPENAI_API_KEY`
`--exa-key <key>`	Exa AI API key for web search in social posts	`$EXA_API_KEY`
`--once`	Process a single video (or next arrival) and exit	Off
`--brand <path>`	Path to `brand.json` config file	`$BRAND_PATH` or `./brand.json`
`-v, --verbose`	Enable debug-level logging	Off
`-V, --version`	Print version and exit	—

Skip Flags

Disable individual pipeline stages:

Flag	Skips
`--no-git`	Git commit/push after processing
`--no-silence-removal`	Dead-silence detection and removal
`--no-shorts`	Short clip extraction
`--no-social`	Social media post generation
`--no-medium-clips`	Medium clip (1–3 min) extraction
`--no-captions`	Caption generation and burning
`--no-social-publish`	Social media queue-build stage

Additional Flags

Flag	Description
`--late-api-key <key>`	Override Late API key

Examples:

bash

# Process without generating shorts or social posts
vidpipe --no-shorts --no-social /path/to/video.mp4

# Skip git (useful during testing)
vidpipe --no-git --watch-dir ./watch

# Transcription + summary only (skip everything optional)
vidpipe \
  --no-silence-removal \
  --no-shorts \
  --no-social \
  --no-captions \
  --no-git \
  /path/to/video.mp4

Environment Variables

Set these in your shell or in a .env file in the working directory.

Variable	Required	Description	Default
`OPENAI_API_KEY`	✅	OpenAI API key for Whisper transcription (and agents when `LLM_PROVIDER=openai`)	—
`LLM_PROVIDER`	❌	LLM provider: `copilot`, `openai`, or `claude`	`copilot`
`LLM_MODEL`	❌	Override the default model for the selected provider	Provider default
`ANTHROPIC_API_KEY`	❌	Anthropic API key (required when `LLM_PROVIDER=claude`)	—
`WATCH_FOLDER`	❌	Directory to monitor for new video files	`./watch`
`OUTPUT_DIR`	❌	Base output directory for processed videos	`./recordings`
`REPO_ROOT`	❌	Repository root for git operations	Current working directory
`FFMPEG_PATH`	❌	Absolute path to `ffmpeg` binary	`ffmpeg` (from PATH)
`FFPROBE_PATH`	❌	Absolute path to `ffprobe` binary	`ffprobe` (from PATH)
`EXA_API_KEY`	❌	Exa AI API key for web search in social posts	—
`BRAND_PATH`	❌	Path to `brand.json`	`./brand.json`

Variable	Required	Description	Default
`LATE_API_KEY`	❌	Late API key for social media publishing	—
`LATE_PROFILE_ID`	❌	Late profile ID (auto-detected if not set)	—

LLM Provider

VidPipe supports multiple LLM providers for AI agent features. Configure via environment variables:

Variable	Required	Description	Default
`LLM_PROVIDER`	❌	LLM provider to use: `copilot`, `openai`, or `claude`	`copilot`
`LLM_MODEL`	❌	Override the default model for the selected provider	Provider default
`ANTHROPIC_API_KEY`	❌	Anthropic API key (required when `LLM_PROVIDER=claude`)	—

Per-Provider Setup

Copilot (default): No extra config needed — uses your GitHub Copilot subscription. Requires an active GitHub Copilot subscription.
OpenAI: Set LLM_PROVIDER=openai. Uses the same OPENAI_API_KEY already required for Whisper transcription.
Claude: Set LLM_PROVIDER=claude and ANTHROPIC_API_KEY=sk-ant-.... Get a key at console.anthropic.com.

Cost Tracking

The pipeline automatically tracks token usage and estimated cost for every LLM call. At the end of each run, a summary is printed showing total tokens, cost (USD for OpenAI/Claude, premium requests for Copilot), and breakdowns by provider, agent, and model. No configuration is needed — cost tracking is always on.

Example `.env` file

env

OPENAI_API_KEY=sk-your-api-key-here
WATCH_FOLDER=/home/you/Videos/Recordings
OUTPUT_DIR=/home/you/Content/processed
REPO_ROOT=/home/you/repos/vidpipe

# Optional: explicit FFmpeg paths
# FFMPEG_PATH=/usr/local/bin/ffmpeg
# FFPROBE_PATH=/usr/local/bin/ffprobe

# Optional: Exa AI for web search links in social posts
# EXA_API_KEY=your-exa-api-key-here

A .env.example file is included in the repository — copy it to get started:
bash
cp .env.example .env

Brand Customization

The brand.json file controls the visual identity and voice of all generated content — captions, social media posts, blog posts, summaries, and short clip descriptions. Customize it to match your personal or company brand.

Location

Place brand.json in your project root. The tool looks for it in the current working directory by default. Override the path with:

CLI flag: --brand /path/to/brand.json
Environment variable: BRAND_PATH=/path/to/brand.json

If no brand.json exists, sensible defaults are used automatically (name: "Creator", handle: "@creator", neutral professional tone).

Example brand.json

json

{
  "name": "Your Name",
  "handle": "@yourhandle",
  "tagline": "Your tagline here",
  "voice": {
    "tone": "professional, friendly",
    "personality": "A knowledgeable content creator.",
    "style": "Clear and concise."
  },
  "advocacy": {
    "primary": ["Technology A", "Technology B"],
    "interests": ["Topic 1", "Topic 2"],
    "avoids": ["Negative comparisons", "Overly salesy language"]
  },
  "customVocabulary": [
    "ProperNoun",
    "TechTermThatWhisperMightMisspell"
  ],
  "hashtags": {
    "always": ["#AlwaysInclude"],
    "preferred": ["#Often", "#Used"],
    "platforms": {
      "tiktok": ["#TechTok"],
      "linkedin": ["#Innovation"],
      "instagram": ["#CodeLife"]
    }
  },
  "contentGuidelines": {
    "shortsFocus": "Highlight key moments and insights.",
    "blogFocus": "Educational and informative content.",
    "socialFocus": "Engaging and authentic posts."
  }
}

Field Descriptions

Field	Type	Description	Default
`name`	string	Your display name — used in content attribution	`"Creator"`
`handle`	string	Social media handle — included in generated posts	`"@creator"`
`tagline`	string	Short bio/tagline for intros	`""`
`voice.tone`	string	Comma-separated tone descriptors for AI writing style	`"professional, friendly"`
`voice.personality`	string	Description of your public persona	`"A knowledgeable content creator."`
`voice.style`	string	How generated content should read	`"Clear and concise."`
`advocacy.primary`	string[]	Core technologies/brands you champion	`[]`
`advocacy.interests`	string[]	Broader topics the AI can reference	`[]`
`advocacy.avoids`	string[]	Things the AI should never include	`[]`
`customVocabulary`	string[]	Proper nouns and jargon sent to Whisper as a prompt hint to improve transcription accuracy	`[]`
`hashtags.always`	string[]	Included on every post, every platform	`[]`
`hashtags.preferred`	string[]	Commonly used — AI picks the most relevant	`[]`
`hashtags.platforms`	object	Platform-specific hashtags (keys: `tiktok`, `youtube`, `instagram`, `linkedin`, `x`)	`{}`
`contentGuidelines.shortsFocus`	string	What moments to extract as short clips	`"Highlight key moments and insights."`
`contentGuidelines.blogFocus`	string	Blog post structure and angle	`"Educational and informative content."`
`contentGuidelines.socialFocus`	string	Social media writing strategy	`"Engaging and authentic posts."`

For full examples (developer, corporate, educator templates) and additional tips, see the Brand Customization Guide.

Output Directory Structure

The --output-dir (default ./recordings) is the base directory. Each video creates a subdirectory named after a slugified version of the original filename:

<output-dir>/
└── <video-slug>/
    ├── <video-slug>.mp4              # Original video copy
    ├── <video-slug>-edited.mp4       # After silence removal
    ├── <video-slug>-captioned.mp4    # With burned-in captions
    ├── README.md                     # AI summary with screenshots
    ├── transcript.json               # Word-level transcript
    ├── transcript-edited.json        # Adjusted transcript (post-silence-removal)
    ├── blog-post.md                  # Long-form blog post
    ├── thumbnails/
    │   └── snapshot-*.png            # Key-moment screenshots
    ├── shorts/
    │   ├── <short-slug>.mp4          # Extracted short clips
    │   ├── <short-slug>-captioned.mp4
    │   ├── <short-slug>-portrait.mp4 # 9:16 platform variant
    │   ├── <short-slug>.ass          # Caption file
    │   └── <short-slug>.md           # Clip description & metadata
    ├── medium-clips/
    │   ├── <clip-slug>.mp4           # 1–3 minute topic clips
    │   ├── <clip-slug>-captioned.mp4
    │   ├── <clip-slug>.ass
    │   └── <clip-slug>.md
    ├── chapters/
    │   ├── chapters.json             # Canonical chapter data
    │   ├── chapters-youtube.txt      # YouTube description timestamps
    │   ├── chapters.md               # Markdown table
    │   └── chapters.ffmetadata       # FFmpeg metadata format
    └── social-posts/
        ├── tiktok.md
        ├── youtube.md
        ├── instagram.md
        ├── linkedin.md
        └── x.md

Common Configurations

Content creator (full pipeline)

bash

vidpipe \
  --watch-dir ~/Videos/Recordings \
  --output-dir ~/Content/processed \
  --brand ./brand.json \
  --verbose

Quick transcription only

bash

vidpipe \
  --no-silence-removal \
  --no-shorts \
  --no-social \
  --no-captions \
  --no-git \
  /path/to/meeting.mp4

CI/CD or automation (no interactive, no git)

bash

OPENAI_API_KEY=sk-... vidpipe \
  --once \
  --no-git \
  --output-dir /tmp/output \
  /path/to/video.mp4

Schedule Configuration

The schedule.json file defines when social media posts are published. It is generated automatically by the pipeline and can be managed via vidpipe schedule. For full details on scheduling and the review workflow, see the Social Publishing Guide.

Troubleshooting

"Missing required: OPENAI_API_KEY"

The tool requires an OpenAI API key. Provide it via:

--openai-key sk-... flag
OPENAI_API_KEY environment variable
.env file in the working directory

"ffmpeg: command not found"

FFmpeg is not on your system PATH. Either:

Install FFmpeg (see FFmpeg Setup)
Set FFMPEG_PATH and FFPROBE_PATH to the absolute paths of the binaries

Verbose mode shows too much output

Verbose mode (-v) sets the log level to debug. If you only need it temporarily, pass the flag on the command line rather than setting it in .env.

Watch mode doesn't detect files

Ensure the --watch-dir path exists and is writable. The watcher monitors for new .mp4 files only. Files that already exist when the watcher starts are not processed — only newly created files trigger the pipeline.

Configuration ​

The Simple Version ​

CLI Parameters ​

Positional Argument ​

Options ​

Skip Flags ​

Additional Flags ​

Environment Variables ​

Social Publishing ​

LLM Provider ​

Per-Provider Setup ​

Cost Tracking ​

Example .env file ​

Brand Customization ​

Location ​

Example brand.json ​

Field Descriptions ​

Output Directory Structure ​

Common Configurations ​

Content creator (full pipeline) ​

Quick transcription only ​

CI/CD or automation (no interactive, no git) ​

Schedule Configuration ​

Troubleshooting ​

"Missing required: OPENAI_API_KEY" ​

"ffmpeg: command not found" ​

Verbose mode shows too much output ​

Watch mode doesn't detect files ​

Configuration

The Simple Version

CLI Parameters

Positional Argument

Options

Skip Flags

Additional Flags

Environment Variables

Social Publishing

LLM Provider

Per-Provider Setup

Cost Tracking

Example `.env` file

Brand Customization

Location

Example brand.json

Field Descriptions

Output Directory Structure

Common Configurations

Content creator (full pipeline)

Quick transcription only

CI/CD or automation (no interactive, no git)

Schedule Configuration

Troubleshooting

"Missing required: OPENAI_API_KEY"

"ffmpeg: command not found"

Verbose mode shows too much output

Watch mode doesn't detect files