Cheapest Veo 3.1 API in 2026: Every Provider's Real Price

Search for the cheapest Veo 3.1 API and most roundups land you on $0.40 per second. That's Google's Standard tier on the Gemini API, and it's the most expensive way to call the model.

If you don't need audio, 4K, or precise first/last-frame control, the same Veo 3.1 weights run for under 5 cents per 8-second 1080p clip on third-party gateways. Roughly a 95% cut. Google opened Veo 3.1 to the Gemini API in October 2025 and added Veo 3.1 Lite on March 31, 2026^[6]. Below is what every public tier actually costs in May 2026, with citations to each provider's pricing page.

TL;DR

• Google Gemini API direct (audio bundled): $0.05/sec (Lite 720p), $0.10–$0.12/sec (Fast 720p/1080p), $0.40/sec (Standard 1080p), $0.60/sec (Standard 4K)^[1].
• fal.ai (audio toggleable): $0.10/sec (Fast 1080p no audio), $0.20/sec (Standard 1080p no audio)^[2].
• Replicate: Veo 3.1 Fast at $0.10/sec without audio^[3].
• reAPI per-generation tier: $0.046 flat for an 8-second 1080p Lite clip, $0.092 flat for an 8-second 1080p Fast clip. Lowest rates I could find anywhere in May 2026^[4].
• The cheap rate forces a fixed 8-second clip with no audio. If you need audio, 4 or 6-second cuts, or first/last-frame control, you'll pay per second instead. The per-second tiers still beat Google's Standard rate by 20–60% on most providers.

How Google prices Veo 3.1

Google bills per second of generated video. Three things change the rate: which model tier you pick (Standard, Fast, or Lite), the output resolution (720p, 1080p, or 4K), and whether audio is bundled. On Google direct, audio is always bundled; on most gateways, it's a toggle.

Google's per-second rates from the Gemini API docs^[1]:

Audio is bundled in every cell. You can't strip it out to save money on Google direct. There's no free tier either; every successful generation gets billed from second one.

Eight seconds is what most users actually generate. At that length:

• Standard 1080p with audio: $3.20 per video
• Fast 1080p with audio: $0.96 per video
• Lite 720p with audio: $0.40 per video

That's the floor going through Google. The third-party gateways are cheaper.

What gateways charge

fal.ai

fal.ai breaks the price down by tier and an audio on/off toggle^[2]:

Stripping audio saves 50% on Standard and about 33% on Fast. Google doesn't expose that control.

Replicate

Replicate charges $0.10/sec for Veo 3.1 Fast without audio^[3]. Same as fal.ai's no-audio Fast rate. The Standard tier and 4K cells aren't listed publicly on Replicate.

OpenRouter

OpenRouter shows "from $0.40/sec" and routes the call to whatever underlying provider it picks^[5]. The floor lines up with Google's Standard tier.

reAPI

reAPI runs two billing modes side by side^[4].

Per-generation tier — flat price for one 8-second clip, no audio:

Per-second tier — same billing model as Google, with first/last-frame control and optional audio:

$0.046 for an 8-second 1080p Lite clip on the per-generation tier is the lowest Veo 3.1 rate I could find publicly listed anywhere in May 2026.

What 1,000 videos a month costs

Take a workload that's actually common: 1,000 eight-second 1080p clips, no audio. Think ads, social loops, B-roll for voice-over.

If you can live with a fixed 8-second clip and no audio, the per-generation tiers run 88–95% cheaper than the next-cheapest option.

When you'll need the per-second tier

The per-generation rate has limits. If any of these matter, price on per-second:

• Audio. Per-generation tiers don't include synthesized audio. A Fast Official 1080p run with audio comes out to $0.138 × 8 = $1.10 — 73% below Google Standard, 8% below fal.ai's Fast with audio, but ~15% above Google Fast.
• Variable duration. Per-generation locks at 8 seconds. Per-second exposes 4 / 6 / 8 second outputs.
• First/last-frame interpolation. Per-generation Fast supports up to 3 reference frames for image-to-video. First-frame plus last-frame anchoring is per-second only.
• 4K resolution. Per-generation 4K cells exist at $0.138 (Lite), $0.276 (Fast), $2.21 (Quality). For premium-quality 4K with audio, the per-second 4K cell at $0.322/s × 8 = $2.58 is the comparable choice.
• Negative prompts, seed, sample_count, person_generation. Per-second only.

Even on the per-second side, reAPI's Fast Official no-audio cell at $0.092/sec is 8% cheaper than fal.ai or Replicate, and 23% cheaper than Google's $0.12/sec 1080p Fast (which forces you to pay for audio).

Faceless-channel math

Picture a daily 90-second video stitched from twelve 8-second 1080p clips, audio dubbed in post. That's 360 clips a month:

• Google Gemini API Fast: 360 × $0.96 = $345.60 / month
• fal.ai Fast (no audio): 360 × $0.80 = $288.00 / month
• reAPI Fast (per-generation): 360 × $0.092 = $33.12 / month
• reAPI Lite (per-generation, 720p): 360 × $0.046 = $16.56 / month

reAPI Lite is 95% cheaper than Google direct for this exact workload. The 720p ceiling stings less than it sounds: Shorts, Reels, and TikTok all re-encode aggressively at delivery and most strip down to 1080p × 30fps no matter what you fed in. Paying for 4K on short-form vertical content is mostly dead money.

When Google direct still wins

Some workloads still belong on Google's first-party API:

1. You're inside an enterprise contract with negotiated Vertex AI rates.
2. You need data residency guarantees Vertex offers and gateways don't.
3. You're using Veo 3.1 alongside Gemini text models in a single billable account for procurement reasons.
4. You need the latest preview model the same hour Google ships it. Gateways generally catch up within hours, occasionally a day or two.

Calling the cheapest Veo 3.1 API tier

reAPI exposes Veo 3.1 through one OpenAI-compatible endpoint:

curl https://reapi.ai/api/v1/videos/generations \
  -H "Authorization: Bearer rk_live_xxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "veo3.1-lite",
    "prompt": "A dolphin leaping through cobalt ocean waves at sunrise",
    "aspect_ratio": "16:9",
    "resolution": "720p"
  }'

The endpoint returns a task_id immediately; poll GET /api/v1/tasks/{task_id} until status is completed. Full schema and SDK examples in Python, Node.js, and Go are on the Veo 3.1 docs page.

To switch to Fast per-generation, change "model": "veo3.1-lite" to "veo3.1-fast" and bump resolution to 1080p. The request shape is identical across the family.

FAQ

What's the absolute cheapest Veo 3.1 API price in 2026?

$0.046 for an 8-second 1080p Veo 3.1 Lite clip on reAPI's per-generation tier (no audio, fixed duration). I could not find a lower per-second or per-generation listed rate across Google, fal.ai, Replicate, or OpenRouter in May 2026.

Does Veo 3.1 have a free tier?

No. Veo 3.1 is paid-tier only on the Gemini API and Vertex AI. New Google Cloud accounts get a $300 / 90-day credit that can be applied to Veo, which works out to roughly 6,000 seconds of Lite 720p generation. Third-party gateways generally don't offer free Veo generations either.

Why is the per-generation rate so much cheaper than the per-second rate?

Per-generation pricing locks the output to the most common shape: 8 seconds, 1080p or below, no audio. That fixed shape lets the gateway batch and forecast compute differently. You give up flexibility (variable duration, audio, first/last-frame anchoring) for the lower headline price. If your product only ever needs an 8-second clip, per-generation is the cheaper math.

Can I get Veo 3.1 with audio under $0.10/second?

Not from any provider I checked in May 2026. Google Gemini API Fast at $0.10–$0.12/sec for 720p–1080p is the cheapest public per-second rate that bundles audio. reAPI's Fast Official with audio is $0.138/sec; fal.ai's Fast with audio is $0.15/sec.

How does Veo 3.1 Lite compare to Veo 3.1 Fast on quality?

Google's release post says Lite delivers "the same speed" as Fast at "less than 50% of the cost," and supports both Text-to-Video and Image-to-Video at 720p and 1080p^[6]. Lite caps at 1080p (no 4K). Subjective output is about a half-tier below Fast on motion fidelity and physical plausibility. Fine for short loops and B-roll. Less ideal for hero shots.

Do third-party gateways add latency vs Google direct?

Barely. Generation itself takes 60 to 180 seconds for an 8-second clip, so the API hop is noise compared to that. Gateway overhead is single-digit percent of wall-clock time, mostly job queueing rather than network round-trip.

What happens if a Veo 3.1 generation fails?

On Google direct, you're not billed for failed generations. Same on reAPI: failed jobs trigger an automatic credit refund. Confirm the failure semantics on your provider before scaling, because client-side retry loops on partial failures can stack costs if the provider charges for upstream-attempted-but-failed jobs.

Is there a way to mix tiers in one application?

Yes. Default to the cheapest tier that satisfies your output spec, then upgrade per-request when needed. On reAPI, the same endpoint serves all five Veo 3.1 model strings — switching from veo3.1-lite to veo3.1-fast-official is a one-field change, no SDK swap.

Picking a tier in practice

Most product workloads can live with a fixed 8-second clip and no audio. For those, reAPI's per-generation Lite or Fast tier ($0.046–$0.092 per 8-second 1080p clip) runs 88–95% cheaper than Google direct, fal.ai, or Replicate.

If you need audio, variable duration, or first/last-frame anchoring, you're on a per-second tier across all providers. reAPI's $0.092/sec Fast Official no-audio cell is still the cheapest of those.

The Standard tier at $0.40/sec really only makes sense inside an enterprise Vertex contract. The cheapest Veo 3.1 API path for almost every other workload is one of the per-generation tiers above.

References

1. Google. Gemini API pricing — Veo 3.1 per-second rates by tier and resolution. Retrieved May 2026 from ai.google.dev/gemini-api/docs/pricing
2. fal.ai. Veo 3.1 — Text to Video. Retrieved May 2026 from fal.ai/models/fal-ai/veo3.1
3. Replicate. Google Veo 3.1. Retrieved May 2026 from replicate.com/google/veo-3.1
4. reAPI. Veo 3.1 — Model page (live pricing). Retrieved May 2026 from reapi.ai/models/veo3-1
5. OpenRouter. Veo 3.1 — API pricing & providers. Retrieved May 2026 from openrouter.ai/google/veo-3.1
6. Google. Build with Veo 3.1 Lite, our most cost-effective video generation model. The Keyword (Google blog), March 31, 2026. blog.google/innovation-and-ai/technology/ai/veo-3-1-lite

Further reading

• Google Cloud. Vertex AI generative AI pricing. cloud.google.com/vertex-ai/generative-ai/pricing
• Google Cloud. Veo 3.1 — Vertex AI model documentation. docs.cloud.google.com/vertex-ai/.../veo/3-1-generate
• fal.ai. Veo 3.1 Fast — Text to Video. fal.ai/models/fal-ai/veo3.1/fast
• reAPI. Veo 3.1 — API reference. reapi.ai/docs/veo3-1