← All guides

Google Flow AI: start with the shot, not the model name

A practical guide to the workspace, the models inside it, and the decisions worth making before you spend credits.

Google Flow is a creative studio for images and video. Veo is a model family used to generate video, not another name for the studio. If you already like working in Flow, there is no reason to switch just to make a short clip. Compare another workspace when a specific input, model or finishing step is missing from your process.

You have a photograph of a desk lamp. You want a small camera move, a warm pool of light and a few quiet seconds for a product page. Then you open the model menu. Flow, Veo, image models, references: suddenly the simple job looks like a vocabulary test. Start by separating the place where you work from the model doing the rendering.

Flow is the desk; the model is one tool on it

Google Flow is the application. Its current help documentation describes creating clips from text, reference media, and start or end frames. You choose a supported model and settings inside that application. Model access and individual controls can change, so a tutorial for an older version may describe a different menu.

Name or controlWhat it means for the job
FlowThe workspace in which you prepare, generate and organize assets.
VeoA video model family, rather than the entire editing workflow.
Start/end framesImages that specify a starting or ending composition, where supported.
Ingredients / referencesMedia that guides subjects or visual details; not a guarantee of exact reproduction.

For the lamp photograph, use less direction

A reference image already supplies the lamp, table and lighting. Re-describing every visible object can introduce contradictions. Use the text to explain what should move and what should remain unchanged. Keep the first attempt modest: a camera move rather than an elaborate transformation.

Use the supplied lamp photograph as the starting composition.
The lamp and table stay still. Make a slow, short camera slide to the right.
Keep the shade shape, switch position and warm light unchanged.
No people, new objects or generated text. End with a steady hold.
Choose a supported duration and aspect ratio in the interface.

This is an example brief, not a tested output. Google’s video creation guide describes adding frames or ingredients, choosing video preferences and generating. It also recommends clean references and avoiding conflicts between the text and visual inputs. The exact controls available to you depend on the selected workflow.

A good first frame can hide a bad clip

Watch the shade edge and the switch, not just the lighting. Does the lamp acquire a second joint halfway through? Does the shade become wider at the end? A pleasant camera move does not make altered merchandise acceptable. Compare the opening, middle and ending with the source, then watch the full clip for flicker. If the reference cannot support the new viewing angle, reduce the movement rather than asking the model to invent the back of the lamp.

Budget for an accepted shot

Before generating, check the current account allowance, chosen model, duration, output count and any separately charged export or enhancement. Do not treat one prompt as one charge if the interface is set to produce several variations. Keep a simple record: attempts charged, refunds credited, clips accepted and manual finishing time. Credits from different services are not comparable units.

We have not run a paid side-by-side benchmark for this guide. There is no universal cost-per-video figure here: the useful number is what your own acceptable result costs after revisions. Check Google’s current account and credit documentation from Flow rather than buying a plan on the strength of an old screenshot.

When another workspace is worth opening

Stay in Flow if its available models and controls already suit the shot. Try another service when you can name the gap: an unavailable model, a required reference mode, or a different way of organizing your work. Write that requirement down before comparing subscriptions. A second prompt box is not an improvement if you still need the same repairs afterward.

You can inspect Wuwei’s video workspace as a separate option, or prepare a source image. Check the live model list and required input mode first. Wuwei is not Google Flow, and opening its workspace does not guarantee access to the same models, reference controls or account allowance. Copy the brief above manually; this article does not submit a generation or transfer your files.

FAQ

Is Google Flow the same as Veo?

No. Flow is the creative application; Veo is a video model family. Model selection is one part of the workflow.

Does Flow accept a photograph as a starting frame?

Google’s current help describes start/end frames and visual references. Availability varies with the model, workflow and region; check your own interface.

Is Wuwei a free replacement for my Flow allowance?

No such equivalence is promised. Wuwei has its own model availability and usage terms. Inspect the current settings and charges before generating.

Take one approved photograph and one simple camera move. Inspect the result before trying a more complicated scene.

Opens the online workspace. Sign-in, model availability and usage charges depend on your account and chosen model; check the displayed terms before generating.