Google Flow Pricing: Credits, Plans and Cost Per Video Ad

Google Flow pricing works out to $4.99, $19.99, $99.99 or $199.99 a month in the US, buying 200, 1,000, 10,000 or 25,000 Google Flow credits, and every account also gets 50 credits a day on top. What decides your real cost is not the plan price, it is that a Veo 3.1 Quality generation costs 100 credits for everybody, including Ultra subscribers. That single number turns the $19.99 Pro plan into ten 8 second clips a month.

If you run paid social for a DTC brand, that is the sentence you actually needed. Everything below is the arithmetic behind it, plus the parts of the credit system that cost people money quietly.

What a Google Flow credit costs, per plan

Google publishes plan prices in one place and credit costs in another, and never multiplies them together. Do it yourself and the picture gets a lot clearer. Divide the monthly price by the monthly allowance:

There is a three-to-one spread between the cheapest and dearest credit. That matters more than it looks, because the plans are not really tiers of the same product. They are volume discounts, and you only claim the discount if you spend the whole allowance.

Google Flow credits: what each generation charges

Google Flow Help publishes the per-generation rates, and they are flat per generation rather than per second:

Two things in that table are worth money.

The charge does not scale with length, so never generate short. Lite and Fast cost the same credits for a 4 second clip as for an 8 second one. If you ask for 4 seconds you have paid for 8 and thrown half away. Generate the full 8 and trim it in the edit, every time. Over a month of drafts that habit alone is worth a meaningful share of your allowance.

The Ultra discount skips the model you ship. Ultra halves Lite and Fast, 10 down to 5 and 20 down to 10. On Quality it is 100 credits for everyone. So the advertised Ultra saving applies to the two tiers you would use for rough drafts, and not to the one you would put spend behind. Ultra is still cheaper per Quality clip, but that comes from the cheaper credit unit price, not from the discount Google names.

Is Google Flow free, really?

There is a free allowance, and the arithmetic quietly disqualifies it for finished work. Every account gets 50 Flow credits a day, subscribed or not, held as one shared balance. Unused daily credits do not roll over. A Quality generation costs 100.

Fifty credits that reset every night can never reach one hundred. With no subscription topping the balance up, the free tier cannot produce a single Veo 3.1 Quality clip, ever. It is not a limited version of the good model. It is locked out of it. What it does reach is two Fast generations a day, or five Lite, which is fine for checking whether the model understands your brief and useless for judging whether the output can carry a campaign.

That is the case for not evaluating AI video on a free tier at all. You will be looking at the quality of a draft model and drawing conclusions about a model you never saw.

Cost per finished video ad, which is the number you are actually buying

Clip prices are not ad prices. Quality tops out at 8 seconds, so a 30 second spot is about four clips stitched together. And nobody ships the first take of anything. Allow three takes a shot, which is modest for ad work, and one finished 30 second ad costs twelve generations at 100 credits each: 1,200 credits.

Look at the Pro row twice, because it is the plan most people land on. At $19.99 with 1,000 credits, one realistically-shot 30 second ad costs 1,200 credits. A single ad costs more than the entire month, and the credits that got you partway there expire at renewal rather than carrying forward. Pro is a plan for finding out what the model can do. It is not a plan for supplying a media buyer who needs creative every week.

The first tier that supports an actual testing cadence is Ultra at $99.99, which is where the real comparison starts. The same arithmetic applied to the underlying model rate card, including the Gemini API route and its per-second pricing, is laid out in full on our Veo 3 pricing breakdown.

The parts of the bill that are not credits

Every number above covers generation only. None of it covers the work that turns generations into an ad:

• Writing a hook that survives the first two seconds of a feed.
• Casting a face that suits the product and staying with it across four separate generations.
• Holding the product itself consistent shot to shot, which is the single hardest thing about assembling AI video into something a brand will sign off.
• Cutting the clips together, adding captions, and cutting the vertical, square and horizontal versions.
• Producing the eight or nine variants a genuine creative test needs, rather than the one hero you fell in love with.

Budget an afternoon per finished ad on top of the credits, and more the first few times. That is not a criticism of Flow. Flow is a shot generator and it is a good one. It just is not an ad tool, and people buy it expecting one.

How to plan a month of credits without running dry

The practical failure mode is not overspending, it is misallocating early. Credits do not roll over, so an unspent allowance is money burned, and an allowance spent on drafts in week one leaves nothing for the version that actually ships.

A sequence that works: block out the concepts on paper first, use Lite at 10 credits to check framing and pacing, move to Fast at 20 for the versions you are choosing between, and reserve Quality for shots that have already survived both. On Ultra at $99.99 that discipline gets you a lot further than ten thousand credits sounds like it should.

Agencies have a harder version of the same problem, because one credit pool has to serve several brands with competing launch dates. At that point which client gets the Quality generations this month stops being a creative decision and becomes a capacity one, and it belongs alongside the rest of your resource planning across competing projects rather than in a creative director's head. The brands that handle this well decide allocation at the start of the cycle, not on the twenty-sixth when the balance is nearly gone.

When to buy top-up credits, and the catch

Google lets Plus, Pro and Ultra subscribers buy top-up AI credits, which are usable in Flow. They do not behave like the monthly allowance. Google notes that unlike monthly Flow credits, purchased AI credits may expire after a period set out when you acquire them. So stocking up ahead of a launch carries an expiry risk your monthly allowance does not, and buying a large balance speculatively is a worse idea than it looks.

Buy top-ups to finish a specific campaign that is already in flight. Do not buy them as a reserve.

Is a Google Flow subscription the right purchase at all?

It depends entirely on what you are producing. If you are generating video and you can direct a model with a written prompt, Flow on Ultra is close to unbeatable on raw price, and nothing that wraps it will match four cents a clip. Go direct.

If what you need is finished ads, the per-clip price stops being the thing to optimize, because the expensive part was never the generation. It was the hour afterwards, multiplied by the number of variants a real test requires. That is the case for buying in finished ads rather than model seconds, which is how UGCGen meters, and how several others in the category do too. We put the whole category's pricing models side by side in our AI UGC pricing breakdown, and if you are weighing a reseller that puts Veo 3.1 behind its own credit system, the Higgsfield pricing teardown shows what that markup does to the per-second cost.

The honest summary: Flow is priced well for people making shots and priced badly for people making ads. Work out which one you are before you pick a tier, because the gap between the $19.99 plan and what a working ad schedule needs is not small, and the plan page will not tell you.