Google Flow for Beginners: Make Your First AI Video

What is Google Flow, in one line?

Flow is a website from Google (labs.google/fx/tools/flow) where you create short AI video clips from text, from images, or from both, and then join them into a longer video. It uses Google's Veo video models for video and Nano Banana for images.

Think of it like a film set in your browser. You are the director. Your prompt is the instruction you give the camera team.

Step 1: Open Flow and start a project

1. Go to labs.google/fx/tools/flow and sign in with your Google account.
2. Click New project. Every project keeps its own clips, images and characters together.
3. Pick 9:16 (vertical) if the video is for Instagram Reels or YouTube Shorts, 16:9 for YouTube.

Flow works on credits. Each generation uses some credits, so plan your shot before you press generate. Free and paid Google AI plans give different amounts; check what your account shows.

Step 2: Write your first prompt (the 5-part formula)

Most beginners write "a man walking" and get a boring clip. A good prompt answers five simple questions:

• Subject: who or what is on screen (a 30-year-old Telugu shop owner, a cup of filter coffee).
• Camera: how we see it (close-up, slow push-in, handheld, top view).
• Action: what happens in these 8 seconds (pours coffee, looks up and smiles).
• Light & location: where and what time (small tiffin centre in Hyderabad, warm morning light).
• Dialogue or sound: what we hear (he says "Good morning!" in Telugu; steel tumbler clink).

Close-up, slow push-in. A 30-year-old Telugu tiffin centre owner in a cotton shirt pours hot filter coffee from a steel tumbler into a davara, steam rising. Warm early-morning light, busy Hyderabad street softly blurred behind him. He looks at the camera and says in Telugu: "Mana coffee taste chesara?" Sound of the street and the coffee pour. No on-screen text, no subtitles.

Step 3: Generate, compare, keep the best

Generate two versions of the same prompt. AI is like a camera team doing two takes: one is usually better. Download only the clips you'll actually use.

If the clip is wrong, change one thing at a time. Wrong camera angle? Edit only the Camera part of the prompt. This is how you learn what each word does.

Step 4: Keep the same person in every shot

The biggest beginner problem: the person's face changes in every clip. The fix is to start from an image instead of only text.

1. First make one clear image of your character (front-facing, good light) with Nano Banana inside Flow.
2. Save that person as a character (or use the image as an ingredient / first frame) in your project.
3. Use that same character for every next shot, and describe the same clothes every time.

Step 5: Join clips into a reel

Each clip is short (around 8 seconds). A reel is 4–6 clips that tell one small story: a hook, the problem, the solution, and the ending with your call to action.

You can arrange clips in Flow's scene builder, or download them and edit in CapCut or any editor you know: trim the dead start and end of each clip, add captions, and export in 9:16.

Common beginner mistakes

• Writing very long prompts with ten actions. One clip = one action.
• Not saying the camera angle, so every clip looks the same.
• Changing the character's clothes by accident between shots.
• Letting AI write text on screen (spelling errors, wrong prices).
• Using every clip AI gives. Choose; don't collect.

Frequently asked questions