Google Flow Skill

Google Flow Skill is a Codex skill for automating Google Flow with a persistent browser session.

It follows the same architecture as notebooklm-skill:

• one bootstrap runner
• one authentication manager
• one local project library
• one task script per workflow

This skill can:

• keep a reusable Flow login session
• sync recent Flow projects into a local library
• activate and search projects
• generate images and videos in an existing Flow project
• upload ingredient images or start/end frames
• list image and video asset edit URLs inside a project
• edit existing Flow images with crop, selection masks, draw masks, text overlays, and prompt-based revisions

Navigation

• Requirements
• Quick Start
• Implemented Workflows
• Generate Images
• Generate Videos
• Edit Images
• Public Repo Notes
• Current Limitations

Implemented Workflows

Workflow	Status	Notes
Flow authentication	Ready	Persistent browser session stored locally
Recent project sync	Ready	Scrapes recent projects from the signed-in Flow home page
Local project activation/search	Ready	Keeps a local project library for repeat runs
Image generation	Ready	Supports model, aspect ratio, and output count
Video generation	Ready	Supports ingredients and frames modes
Ingredient uploads	Ready	Uses absolute local file paths
Existing image edits	Ready	Full edit, select-box, draw-rect, draw-brush, crop, text
SceneBuilder automation	Not included	Still out of scope

Requirements

• Windows with Python 3.10+
• Google Chrome installed
• A Google account with access to Flow

Python dependencies:

• patchright==1.55.2
• python-dotenv==1.0.1

Layout

google-flow/
├── SKILL.md
├── README.md
├── README.zh-CN.md
├── agents/
├── references/
└── scripts/

Key scripts:

• scripts/run.py: environment bootstrap and entrypoint
• scripts/auth_manager.py: persistent login state
• scripts/project_manager.py: local project library and asset listing
• scripts/generate_media.py: image and video generation
• scripts/edit_image.py: image editing workflows
• scripts/cleanup_manager.py: local cleanup helper

Quick Start

1. Check authentication:

python scripts/run.py auth_manager.py status

2. If needed, sign in to Flow:

python scripts/run.py auth_manager.py setup

3. Sync recent projects from the signed-in Flow home page:

python scripts/run.py project_manager.py sync
python scripts/run.py project_manager.py list

4. Activate a project:

python scripts/run.py project_manager.py activate --id PROJECT_ID

5. Generate or edit media inside that project.

Authentication

The skill stores browser state in data/browser_state/ and reuses it on later runs.

Useful commands:

python scripts/run.py auth_manager.py status
python scripts/run.py auth_manager.py setup
python scripts/run.py auth_manager.py validate
python scripts/run.py auth_manager.py reauth
python scripts/run.py auth_manager.py clear

Project Management

Sync, search, activate, and inspect projects:

python scripts/run.py project_manager.py sync
python scripts/run.py project_manager.py list
python scripts/run.py project_manager.py search --query storyboard
python scripts/run.py project_manager.py activate --id PROJECT_ID
python scripts/run.py project_manager.py assets --id PROJECT_ID --kind image
python scripts/run.py project_manager.py assets --id PROJECT_ID --kind video

Generate Images

Create one portrait image:

python scripts/run.py generate_media.py \
  --project-id PROJECT_ID \
  --mode image \
  --prompt "A small dog dancing joyfully, Pixar-style 3D character, warm cinematic light" \
  --model "Nano Banana 2" \
  --aspect-ratio portrait \
  --outputs 1

Generate Videos

Generate a video from text:

python scripts/run.py generate_media.py \
  --project-id PROJECT_ID \
  --mode video \
  --video-mode ingredients \
  --prompt "A handheld tracking shot through a rainy neon alley" \
  --model "Veo 3.1 - Fast" \
  --aspect-ratio landscape \
  --outputs 1

Generate a video from frames:

python scripts/run.py generate_media.py \
  --project-id PROJECT_ID \
  --mode video \
  --video-mode frames \
  --prompt "The same character moves naturally with smooth cinematic motion" \
  --start-frame "C:/absolute/path/start.png" \
  --end-frame "C:/absolute/path/end.png" \
  --model "Veo 3.1 - Fast" \
  --aspect-ratio portrait \
  --outputs 1

Edit Images

List available image edit URLs first:

python scripts/run.py project_manager.py assets --id PROJECT_ID --kind image

Full-image prompt edit:

python scripts/run.py edit_image.py \
  --asset-url "https://labs.google/fx/zh/tools/flow/project/.../edit/..." \
  --tool full \
  --prompt "Turn this into a moody dusk scene"

Selection mask edit:

python scripts/run.py edit_image.py \
  --asset-url "https://labs.google/fx/zh/tools/flow/project/.../edit/..." \
  --tool select-box \
  --box "0.18,0.20,0.55,0.48" \
  --prompt "Replace this area with a rainy city skyline"

Rectangle mask edit:

python scripts/run.py edit_image.py \
  --asset-url "https://labs.google/fx/zh/tools/flow/project/.../edit/..." \
  --tool draw-rect \
  --box "0.12,0.58,0.38,0.86" \
  --prompt "Remove the foreground object"

Brush mask edit:

python scripts/run.py edit_image.py \
  --asset-url "https://labs.google/fx/zh/tools/flow/project/.../edit/..." \
  --tool draw-brush \
  --points "0.20,0.62;0.26,0.64;0.31,0.69;0.36,0.75" \
  --prompt "Erase this object and blend the background naturally"

Text overlay:

python scripts/run.py edit_image.py \
  --asset-url "https://labs.google/fx/zh/tools/flow/project/.../edit/..." \
  --tool text \
  --point "0.30,0.40" \
  --text-size 24 \
  --text "OPENING SOON"

Crop:

python scripts/run.py edit_image.py \
  --asset-url "https://labs.google/fx/zh/tools/flow/project/.../edit/..." \
  --tool crop \
  --crop-preset square \
  --box "0.12,0.18,0.82,0.78"

Coordinate rules:

• --box and --points use normalized values from 0 to 1
• --point uses the same normalized coordinate system
• --text should be single-line
• --text-size maps to Flow's text-size slider in px

Notes

• Always run scripts through python scripts/run.py ...
• Prefer syncing recent projects before adding project URLs manually
• Use absolute file paths for ingredient images and video frames
• Generation and prompt-based edits may consume Flow credits

Public Repo Notes

This public repository contains only the skill source code and documentation.

It does not publish:

• browser login state
• Flow cookies or session files
• downloaded media assets
• local virtual environments
• .env files

Those runtime files are excluded by .gitignore, especially data/, .venv/, and .env.

Current Limitations

• SceneBuilder automation is not included
• Download automation is not included
• Flow UI labels may vary by locale and can require selector updates

References

• Internal workflow notes: references/workflow.md
• Official Flow landing page: https://labs.google/fx/tools/flow
• Flow video help: https://support.google.com/labs/answer/16353334?hl=en
• Flow image help: https://support.google.com/labs/answer/16729550?hl=en