Google Flow Labs Skill

What is Google Flow?

Google Flow (accessible at flow.google, developed under Google Labs) is a unified AI creative studio for generating and editing images and videos. It launched originally as VideoFX in May 2024 and was relaunched as a full creative workspace on February 25, 2026.

Flow merged three previously separate Google Labs tools into one interface:

Flow (original AI video generator)
Whisk (visual collage and mood board tool)
ImageFX (text-to-image generator)

Starting March 2026, users can opt in to automatically transfer legacy projects from Whisk and ImageFX directly into the new Flow library.

Underlying Models

Flow is powered by three core AI models:

Model	Technical name	Role
Nano Banana 2	Gemini 3.1 Flash Image	Fast image generation, 14 reference inputs, 512px preview mode, extreme aspect ratios (4:1, 8:1), live web search grounding
Nano Banana Pro	Gemini 3 Pro Image	Higher-fidelity image generation, thinking mode with interim thought images, 14 reference inputs with slot fidelity tiers (slots 1–6 high-fidelity, 7–14 supplementary)
Veo 3.1	veo-3.1-generate / veo-3.1-fast-generate-preview	Text-to-video and image-to-video. Native audio: environmental sounds, ambient noise, and broad speech rhythm. Note: exact frame-level lip sync is not guaranteed — plan VO alignment in post for dialogue-critical content.
Gemini	—	Powers natural language prompting and planned multi-shot scene breakdown (announced, not yet live)

Flow offers two speed variants for Veo 3.1:

Fast (veo-3.1-fast-generate-preview) — 2x speed, suitable for iteration and UGC. Use at 720p during drafts.
Standard (veo-3.1-generate-preview) — full quality for final renders. Use at 1080p for deliverables.

Some features are only available on specific model tiers. Flow automatically switches to a compatible model if needed.

Core Features

Image Generation

Generate high-fidelity images from text prompts using Nano Banana
Use as standalone images or as "ingredients" for video workflows
Text rendering: wrap text in quotes in the prompt for legible in-scene text (e.g. signs, labels)

Video Generation

Text to Video: generate clips directly from natural language descriptions
Ingredients to Video: use reference images to anchor character/subject consistency throughout a video. Requires Ultra tier. Accepts up to 3 reference images of a single subject — cannot be split across multiple subjects.
First Frame / Last Frame (called "Frames to Video" in legacy docs): use a start image, end image, or both as locked frames. Veo generates the transition between them. Independent of Ingredients slots — both can be used simultaneously.
Extend: extend an existing clip to continue the scene. Clips can be chained to reach ~1–2.5 minutes total.
Camera Control: two separate systems — (1) prompt-time direction via the Cinematography layer of the prompt; (2) Camera Adjustment UI tool applied post-generation to reframe, orbit, or dolly an already-generated clip. System 2 works best on clips without existing camera motion.

Editing

Lasso Tool: draw a selection around any region of an image; apply conversational prompts like "remove the man" or "add koi fish" to that region only
Image + Text Editing: target specific parts of images with prompts without re-generating the full image (use masking instead of full re-rolls for efficiency)
Video Editing: extend clips, control camera motion, make scene adjustments mid-generation

Asset Management

Asset Grid: refreshed UI for tagging, sorting, and grouping media assets
Collections: group related assets by pressing Shift+Click or click-and-drag to select multiple tiles, then create a Collection
@ Reference: use @filename in prompts to reference specific assets from your library
Individual generation tiles (batched generations removed in 2026 redesign)

Access & Pricing

Plan	Cost	Flow Access
Free	$0	Basic access after signup (age 18+ required)
Google AI Pro	$19.99/month	Full Flow experience — Veo 2 + Veo 3.1 models, credits top-ups available
Google AI Ultra	$249.99/month	All Pro benefits + more credits, first access to experimental models, Ingredients to Video
Google AI Ultra for Business	Separate pricing	Ultra tier for business accounts
Workspace plans	Included	Business Starter/Standard/Plus, Enterprise tiers, G Suite Legacy Free, Education Standard/Plus, Google AI Pro for Education — all get 100 monthly credits at no extra charge

Note: Flow is not available in all countries where Google AI Pro is offered. VPNs do not enable access in unsupported regions. Check flow.google for supported regions before subscribing.

Credits

Credit costs vary per model and are shown in-product before confirming generation
Audio Generation Failed errors result in full credit refunds
Additional credit top-ups can be purchased on Pro tier

Technical Notes

Best experience: Desktop computer with a Chromium-based browser (Google Chrome recommended)
Mobile and non-Chromium browsers are not fully optimized as of March 2026
SynthID Watermarks: All outputs (images and videos) contain invisible SynthID watermarks identifying AI-generated content. Google AI Pro users' videos also receive a visible watermark
"Pending" or error cards: check top-right corner for system notifications; try refreshing

Google Labs FX Ecosystem

Flow is the hub of the Google Labs FX platform. Companion tools that integrate with Flow:

MusicFX: generate royalty-free background music synchronized to your visuals
Whisk and ImageFX: now merged directly into Flow (as of Feb 25, 2026)

Workflow pattern: use Flow to generate visual mood boards → feed emotional cues into MusicFX for synchronized audio → complete a full creative pipeline without leaving Google's ecosystem.

Prompting Best Practices

Text in scenes: wrap desired text in double quotes in your prompt — e.g. a billboard reading "SALE 50% OFF" for legible in-scene text
Partial edits: use the lasso tool to mask only the area to change; don't re-roll the whole image if 90% is correct
Asset referencing: use @assetname in prompts to pull specific images from your library into generation
Camera specificity: describe exact camera movements (slow pan left, zoom into subject's face) for cinematic results
Audio-aware prompting: for Veo 3.1, describe ambient sounds, dialogue, and music cues in the prompt for native audio generation

Scale & Adoption (as of Feb 2026)

Over 1.5 billion images and videos created since Flow launched
Over 40 million AI videos generated using Google's Veo models
Available to Google Workspace business, enterprise, and education users

Planned / Announced Features

Gemini-powered multi-shot prompting: describe a complex scene and AI breaks it into multi-shot sequences automatically (announced, not yet live)
YouTube integration: direct-to-YouTube publishing expected before end of 2026
Paid tiers: pricing expected to follow Veo 3 API model (per-second generation charges) once free beta matures

Quick Reference Links

App: https://flow.google
Official help: https://support.google.com/flow/answer/16353333
Google Labs blog post (Feb 25, 2026): https://blog.google/innovation-and-ai/models-and-research/google-labs/flow-updates-february-2026/
Discord community: #flow channel on Google Labs Discord

google-flow-labs