Gflow Cli

What gflow-cli does

gflow-cli exposes Google Flow generation workflows through both a Python command-line interface and an MCP server. It drives Veo for text-to-video, image-to-video, reference-to-video, and clip-extension tasks, and uses Imagen for text-to-image generation. The CLI also supports creating reusable Flow Characters with face and body references, which can help maintain a subject across generations.

The package is intended for scripted, batched, and pipeline-based work. Prompts can be submitted from shell loops, and a creative-director tool can rewrite concise prompts before generation. Generated files are written locally by default, with support for routing output to S3, MinIO, or Google Cloud Storage through the storage configuration.

How it works

Authentication and generation use a persistent Playwright Chromium profile connected to the user's own Google Flow session. The initial login opens a real Chrome window, while later commands reuse the stored session. gflow auth renew can refresh a stale session without requiring the normal interactive flow.

MCP operation is available through gflow mcp run or gflow serve. The browser is headed because Google authentication and reCAPTCHA checks require that mode. The first generation can take 30 to 90 seconds while Chromium starts; subsequent calls reuse the warmed session.

Veo video generation consumes the Flow credits associated with the Google account. Image generation and composition operations are free according to the project’s usage notes. Batch records are saved locally before submission, and the CLI reports distinct failures instead of silently resubmitting after interface changes.

Setup and configuration

Install the package with either of these approaches:

uv tool install gflow-cli
# or
pip install gflow-cli

Install Chromium once with Playwright:

uv tool run --from gflow-cli playwright install chromium

Authenticate the browser profile, then check the session and credit balance:

gflow auth login --browser chrome
gflow auth status
gflow credits user

Set GFLOW_CLI_BROWSER_PROXY when the browser must use a proxy. Generated files go under GFLOW_CLI_OUTPUT_DIR unless GFLOW_CLI_STORAGE_URI routes them to S3, MinIO, or Google Cloud Storage. The package accepts Google accounts with Flow access, and each generation remains subject to that account’s credit allowance.

Tools and capabilities

• Generate images from text with Imagen using gflow image t2i.
• Generate Veo clips from text, images, or references with gflow video t2v, i2v, and r2v.
• Extend an existing video beyond Flow’s eight-second clip limit with gflow video extend.
• Create reusable face-and-body subject references with gflow character create.
• Rewrite prompts with the creative-director tool or supply custom tools through the project’s My Tools system.
• Run generation jobs from shell scripts, batches, and larger content pipelines.

Limitations and notes

gflow-cli is alpha software and reverse-engineered rather than an official Google product. It automates a browser session, so Google’s terms, account policies, interface changes, and authentication behavior remain relevant. The project does not use proxies or fingerprint spoofing to disguise automation, and it cannot guarantee account outcomes.

Google Flow UI changes can break selectors or workflows. Video submissions can spend credits, so batch operations should be tested with care. The browser profile and account session are central to operation; this is not a direct Google API client. Playwright and Chromium must be available, and installations from a local checkout may resolve dependency versions that differ from the project’s tested environment.