Veo 3.1 API Cost: Google Direct vs Resellers

Here is the finding that inverts conventional wisdom: for Google's Veo 3.1, the official Vertex AI API is the cheapest route. Resellers match Google's price at best and charge 50% more at worst — and Google exclusively offers two budget tiers no reseller in our dataset carries. Every rate below verified July 30, 2026.

Google's Tier Ladder vs the Resellers

Google Vertex AI publishes a three-tier ladder (Standard, Fast, Lite) with separate audio and video-only rates. The two cheapest tiers, Fast and Lite, are either cheaper on Google than on resellers or unavailable from resellers entirely.

Sources: Google Vertex AI, fal.ai, WaveSpeedAI

Read the Fast row with audio: Google charges $0.10/s, both resellers charge $0.15/s. That is a 50% reseller markup on the tier most developers actually use. On the Standard tier the resellers match Google exactly ($0.20 / $0.40) — no markup, but no discount either.

Why Resellers Lose Here

The Veo 3.1 spread across our dataset is 20x, from Google's Lite video-only tier at $0.03/s to Google's Standard 4K+audio tier at $0.60/s. Critically, both ends of that spread are Google's own official rates. Resellers occupy only the middle: fal.ai and WaveSpeed cluster around the Fast and Standard tiers and never touch Lite.

That leaves Google owning the price floor. If your workload tolerates Lite quality, $0.03/s video-only is less than a third of the cheapest reseller Veo rate ($0.10/s, fal.ai Fast video-only). There is simply no reseller path to Lite pricing in our dataset. You go direct to Vertex AI or you don't get it.

The reason is structural: Veo runs on Google's own TPUs inside Vertex. Resellers proxy the same API and add a convenience margin (one key, unified billing across models). For a single-vendor model like Veo, that convenience buys you nothing on price — you are paying a markup to avoid a Google Cloud signup.

There is a second structural quirk worth flagging. The prior generation, Veo 3, listed at roughly $0.75/s before its June 30, 2026 retirement (per Google's pricing history; Veo 3 is outside our verified dataset — Google Vertex AI pricing history). Veo 3.1 Standard with audio now sits at $0.40/s in our verified data, and the Lite tier reaches down to $0.03/s. Between generations, Google rebuilt the price ladder downward and added budget tiers that undercut the aggregator layer. That is unusual first-party behavior, and it is the mechanism behind the inverted pricing you see here. Resellers priced Veo 3.1 against the old assumption that first-party APIs sit at the top; Google's Lite and Fast tiers moved the floor out from under them.

Per-Clip Cost Tables

Veo 3.1 caps at 8 seconds per generation (model spec), so clip costs are bounded. Here is what a full 8-second clip costs at each route, plus a common 5-second clip:

Fast tier, 720p, with audio (the workhorse config):

Lite tier, 720p (Google only):

Standard tier, with audio:

At production scale, the Fast-tier markup compounds. Ten thousand 8-second clips at 720p Fast + audio: $8,000 on Google versus $12,000 on fal.ai or WaveSpeed. The $4,000 gap is pure reseller margin for a model you could call directly.

The Audio Caveat: Veo Always Bundles It Differently

One normalization trap sets Veo apart from its competitors. Google prices the audio flag explicitly and non-uniformly across tiers:

Source: Google Vertex AI

Note the surcharge is not a flat percentage — it ranges from +25% (Fast) to +100% (Standard). This matters when you compare Veo against models with a fixed audio penalty. Kling 3.0, for instance, applies a uniform +50% for audio at every tier. Veo's Standard tier doubles in price for audio, so if you are generating muted clips for a post-production pipeline, the Standard video-only rate ($0.20/s) is your real cost — not the $0.40/s headline.

The other side of the caveat: Veo 3.1 includes native audio-visual generation as a first-class feature, and Google's audio is widely rated best-in-class for lip-sync. Competitors like Sora 2 and Hailuo 2.3 carry no audio flag in our dataset (audio support unconfirmed). Compare Veo's video-only rate against no-audio models, and its audio-on rate against native-audio models (Kling, Seedance).

Recommendation by Workload

Cheapest possible Veo (quality-tolerant): Google Vertex Lite, video-only, 720p at $0.03/s. No reseller comes close.

Best-value production tier: Google Vertex Fast + audio at $0.10/s. Native audio, 720p or 1080p ($0.12/s), and a third cheaper than any reseller. This is the default recommendation for most Veo 3.1 builds.

Highest quality: Google Vertex Standard 4K + audio at $0.60/s. Only official path to 4K in our dataset.

If you must use a reseller (existing fal.ai billing relationship, multi-model key): fal.ai carries the fullest Veo ladder and matches Google exactly on Standard ($0.20 / $0.40). You only overpay on the Fast tier. WaveSpeed carries only Fast at $0.15/s.

Bottom line: for Veo 3.1 specifically, going direct to Google is both cheaper and gives you more tiers. Resellers earn their margin on multi-model convenience, not on Veo unit economics. Run your clip volume through /calculator, and check /cheapest/veo-3.1 for live floor prices. Our normalization approach is documented at /methodology.

Prices verified July 30, 2026 and subject to change without notice. Check the live tracker for current rates. Some provider links on this site may be referral links.