Veo 3.1 API Pricing: Fast and Lite Cost per Second, Clip by Clip

A video model price list looks simple until you try to budget with it. Five cents per second sounds like pocket change, then you generate 400 clips, throw away half of them, and the invoice stops matching your mental math. This article puts real numbers on Veo 3.1 API pricing for the Veo 3.1 Fast and Veo 3.1 Lite tiers, with the standard Veo 3.1 model alongside for reference, so you can price a clip, a campaign, or a month of output before you write a line of code.

Every rate below comes from Google's Gemini API pricing page, checked in early October. Wherever I calculate something, the arithmetic is on the page, so you can redo it with your own clip lengths and retry rates.

The Price List at a Glance

Per-Second Rates by Tier

Google bills Veo by the second of generated video. The API has three tiers, and each one carries its own rate for each resolution:

The ladder is easy to remember. At 720p, Fast costs twice as much as Lite, and Standard costs four times as much as Fast, which makes Standard eight times the price of Lite. In the preview naming scheme the three model IDs are veo-3.1-generate-preview, veo-3.1-fast-generate-preview and veo-3.1-lite-generate-preview.

💡 Quick rule: multiply the rate by the clip length. An 8-second Lite clip at 720p is 8 × $0.05 = $0.40.

What the Rate Includes

The per-second number is only part of the story. These details decide what you actually pay:

• Audio is bundled. The pricing page lists audio as included by default, so the Gemini API has no separate sound line item.
• Failed generations are free. Google states that you are only charged when a video is successfully generated, so an audio processing hiccup that blocks a clip does not bill you.
• Paid tier only. These models are not available on the free tier, so you need billing enabled before the first call.
• Lite stops at 1080p. There is no 4K option on Lite, which is the trade for its low rate.
• No input-type surcharge. The page lists rates by tier and resolution, not by text-to-video versus image-to-video.

Lite is the newest tier. It arrived at the start of April, and launch reports described it as costing less than half of what Fast cost at the time, when Fast was listed at $0.15 per second. A Fast price cut was scheduled for April 7, and the page now shows $0.10 at 720p. If you find an older post quoting $0.15 for Fast, that is the reason.

One limit on this article: I could not confirm the Vertex AI rate card, so if you call these models through Google Cloud instead of the Gemini API, check that page rather than assuming the numbers match.

What a Single Clip Costs

4, 6 and 8 Second Clips

Both the Lite and Fast model pages list the same three durations: 4, 6 or 8 seconds. Here is the bill per clip at 720p:

Cost scales in a straight line with length, so a 4-second clip is exactly half of an 8-second one. That matters more than it looks. Most social edits cut between shots every two to four seconds, and if you only use 3 seconds of each generation you still pay for all 8. Asking for a 4-second clip from the start cuts that waste in half.

Resolution Changes the Bill

Resolution adds a surcharge on Fast and Lite, but the size of that surcharge differs a lot between them. On the Lite model page, 1080p requires the 8-second duration, so the fair comparison is the cost of one 8-second clip at each resolution:

Three things stand out:

1. Lite at 1080p ($0.64) is cheaper than Fast at 720p ($0.80). If you want sharper output on a tight budget, resolution on Lite is a better deal than a tier upgrade.
2. Fast adds only 20 percent going from 720p to 1080p. Lite adds 60 percent for the same step.
3. Standard charges the same at 720p and 1080p. If you are paying Standard rates, you may as well ask for 1080p.

Fast vs Lite: Where the Gap Sits

At 720p, Fast Costs Double

For the same 8-second clip at 720p, Fast costs $0.80 and Lite costs $0.40. The real question is whether Fast output is worth twice the budget per take. One early hands-on test reported that an 8-second Lite clip rendered in under a minute and that the quality gap between Lite and Fast was less noticeable than the gap between Fast and Standard.

Speed points the same way. In the example clips on the PicassoIA model pages, 8-second 720p generations from Lite finished in roughly 37 to 42 seconds, while the Fast examples took about 45 to 59 seconds. Those are small samples, so treat them as a rough feel rather than a benchmark.

At 1080p, the Gap Shrinks

At 1080p, Fast costs $0.12 per second and Lite costs $0.08, a difference of four cents per second, or $0.32 on an 8-second clip. If you were going to pay for 1080p anyway, that extra 32 cents buys the Fast tier's extra controls on PicassoIA, including a negative prompt field and an audio on/off switch that the Lite page does not list.

When Standard Still Makes Sense

Standard runs $3.20 per 8-second clip at 720p or 1080p and $4.80 at 4K. That is premium money, so save it for hero shots, client deliverables and moments where a rejected take costs more in time than in dollars. Also note that Fast offers 4K at $0.30 per second, or $2.40 for 8 seconds, which is half of Standard's 4K price. If 4K is the only reason you were eyeing Standard, check Fast first.

A rough way to match tier to job:

Retries Are the Hidden Cost

Google bills for clips that generate successfully. It does not care whether you like them. Every take where a hand looks wrong or the camera move drifts still costs full price, and that is where budgets quietly break.

Cost per Kept Clip

Here is what one kept 8-second clip costs at 720p, depending on how many takes it needed:

Eight Lite takes cost the same as one Standard take. If Lite needs three attempts per keeper and Standard nails it on the first, Lite still wins $1.20 to $3.20. The tier only loses on cost when it needs more than eight times as many takes as Standard.

💡 Track your hit rate. Log how many generations you run for every clip you keep. That one ratio tells you more about your real cost than any rate card.

A 30-Second Ad Budget

A 30-second ad built from four 8-second shots needs 32 seconds of kept footage. Assume three takes per shot, which makes 12 generations:

Even the top row stays under sixty dollars for a finished spot, before editing time. The bottom line is that the model tier is rarely the expensive part of a video job. Ten wasted takes on Standard cost more than the whole Lite budget.

Monthly Budgets for Real Workloads

Three Clips a Day

Picture a small brand posting three 8-second clips per day for 30 days. That is 90 kept clips. At three takes each, you run 270 generations a month:

The gap between Lite at 720p and Standard at 1080p is $756 per month for the same number of clips. Scale the volume up tenfold, to 2,700 takes, and the same four rows read $1,080, $2,160, $2,592 and $8,640. Vertical 9:16 output fits this plan well, since the pricing page does not list a different rate for portrait video.

Draft Cheap, Finish Better

The cheapest workflow mixes tiers. Draft prompts on Lite at 720p, then re-run the winning prompt on Fast at 1080p for the final. Ten Lite drafts cost $4.00 and one Fast 1080p final costs $0.96, a total of $4.96. Three straight Standard takes cost $9.60.

One caveat: the same prompt will not return identical frames on a different tier. Treat the draft as a prompt test, not a preview of the exact shot.

A Small Cost Calculator

Python Snippet for Estimates

If you generate in bulk, put the rates in code and print the estimate before you submit the job:

RATES = {  # USD per second of generated video
    "lite": {"720p": 0.05, "1080p": 0.08},
    "fast": {"720p": 0.10, "1080p": 0.12, "4k": 0.30},
    "standard": {"720p": 0.40, "1080p": 0.40, "4k": 0.60},
}

def estimate(tier, resolution, seconds, clips, takes_per_keeper=3):
    rate = RATES[tier][resolution]
    return rate * seconds * clips * takes_per_keeper

print(estimate("fast", "1080p", 8, 90))  # 259.2

Copy Prompt

Two habits keep that number honest. Refresh RATES whenever Google announces a change, as it did in April. And log your real takes_per_keeper from live runs instead of guessing 3.

Three mistakes inflate the bill more than any rate difference:

• Rendering 8 seconds when you keep 3. Pick the shortest duration your edit needs.
• Paying for 4K by default. Only Fast and Standard offer it, and the 4K rate is 2.5 to 3 times the 1080p rate.
• Skipping the draft stage. Prompts that fail on Lite at $0.20 per 4-second clip would have failed on Standard at $1.60.

How to Use Lite on PicassoIA

If you would rather test prompts without setting up the API first, Veo 3.1 Lite is available on PicassoIA, and its model page describes online use as free with no coding needed.

Step-by-Step Settings

1. Write the prompt. Describe the subject, the action, the camera move and the sound, since audio is generated with the clip.
2. Add a start image (optional). For image-to-video, the ideal size is 1280x720 for 16:9 or 720x1280 for 9:16.
3. Add an end image (optional). With both frames, the model creates a smooth transition between the two stills.
4. Choose the aspect ratio. The options are 16:9 and 9:16.
5. Pick the resolution. 720p is the default, 1080p is available, and 4K is not supported.
6. Set the duration. Choose 4, 6 or 8 seconds. The default is 8, and 1080p requires 8.
7. Set a seed (optional). It improves consistency but does not guarantee identical results.

💡 Budget tip: run your first test at 720p and 4 seconds. At Google's list rate, that draft is worth $0.20.

Fast Adds Extra Controls

Veo 3.1 Fast on PicassoIA defaults to 1080p and adds a negative prompt field plus a Generate Audio switch. Turning audio off is handy when you plan to add a music bed in an editor. Do not assume it lowers your bill, though: Google's pricing page lists audio as included, so check the rate card of wherever you run the model.

Other Models Worth Testing

Veo is not the only option, and a quick side-by-side test can save real money. These are all available on PicassoIA:

• Seedance 2.5 Lite is listed as free and unlimited, with clips up to 10 seconds.
• PicassoIA Video is a free, unlimited generator for text or image input.
• Seedance 2.0 Mini handles text to video with native audio.
• Wan 3 and Kling v3 Video target cinematic motion.
• LTX 2.3 Fast renders text to 4K video quickly.

Run one prompt through two or three of these before you commit a budget to a single model.

Try Your Own Clips Today

A price table only tells you so much. The fastest way to find your own hit rate is to generate. Open Picasso IA, run a draft on Veo 3.1 Lite, send the same prompt through Veo 3.1 Fast, and compare the clips side by side. Count how many takes each one needs, then multiply by the rates above.

You can also start from a still. Create a first frame with a text-to-image model, upload it as the start image, and animate it. Browse every available option at picassoia.com/en/all-models, pick two models, and make your first clip and your first image today.