Google Veo public-data review for creators and small businesses
Google sells Veo through two doors with completely different pricing, and picking the wrong one is the most expensive mistake buyers make. Through the Gemini API you pay per second of finished video, from $0.05/second for Veo 3.1 Lite at 720p up to $0.60/second for 4K, audio included. Through the Flow studio you pay nothing per second — you spend a monthly credit allowance attached to a Google AI subscription, starting at 50 free credits a day. If you are generating a handful of clips a month, Flow is far cheaper; if you are generating programmatically at volume, the API is the only option that scales.
Door one: the Gemini API, priced per second
Google's official Gemini API pricing page lists Veo 3.1 in three variants, all billed per second of output with audio generated by default.
| Variant | 720p | 1080p | 4K |
|---|---|---|---|
| Veo 3.1 Standard | $0.40/sec | $0.40/sec | $0.60/sec |
| Veo 3.1 Fast | $0.10/sec | $0.12/sec | $0.30/sec |
| Veo 3.1 Lite | $0.05/sec | $0.08/sec | Not supported |
Two things are worth reading twice. First, the price is per second of output, not per clip — an eight-second clip at Fast 1080p is about $0.96, while the same clip at Standard 1080p is $3.20, a 3.3x difference for one dropdown. Second, Google's page notes that an audio processing issue can occasionally prevent a video from being generated, and that you are only charged when generation succeeds.
A third-party route is worth pricing against these rates. Pictory resells Veo 3.1 inside its editor at 20 credits per second at 1080p, and Veo 3.1 Fast at 10. Converted at Pictory's cheapest marginal credit rate — about $0.0398 from its largest top-up pack — that is roughly $0.80 and $0.40 a second, against Google's $0.40 and $0.12 above: about 2x and 3.3x list. The markup pays for Pictory's editor, captioning and stock library rather than for model access, so if generation volume is your bottleneck, call the Gemini API directly.
Door two: Flow, priced in credits
Flow is Google's studio interface for Veo, and it does not use per-second billing at all. Your allowance comes from whichever Google AI subscription you hold. Google's Flow credits help page publishes both the allowances and the cost of each generation.
| Plan | Credit allowance | Lite (10 credits) | Fast (20 credits) | Quality (100 credits) |
|---|---|---|---|---|
| No subscription | 50 credits per day | 5 per day | 2 per day | Exceeds a day's allowance |
| Google AI Plus | 200 per month | 20 | 10 | 2 |
| Google AI Pro | 1,000 per month | 100 | 50 | 10 |
| Google AI Ultra ($100) | 10,000 per month | 2,000 | 1,000 | 100 |
| Google AI Ultra ($200) | 25,000 per month | 5,000 | 2,500 | 250 |
The generation counts above are arithmetic from Google's published credit costs, not a separate Google claim. Ultra subscribers get a discount on the cheaper variants — Google lists Lite at 5 credits and Fast at 10 credits for Ultra, which is what doubles the Ultra columns; Quality stays at 100 credits on every plan. One caveat on the free row: Quality costs 100 credits against a 50-credit daily allowance that does not roll over, so treat Quality as effectively out of reach without a subscription and confirm the current rule with Google if that is your plan.
On subscription prices, Google's official announcement of May 19, 2026 introduced a new AI Ultra tier at $100/month and cut the top Ultra tier from $250 to $200/month. AI Pro is $19.99/month. AI Plus was reduced to $4.99/month in June 2026 according to reporting at the time — that figure is press-reported rather than quoted from a Google page here, so verify it at checkout.
What Veo 3.1 actually does
Veo 3.1 has been Google's current video model since its launch on October 15, 2025. Google's DeepMind model page describes text-to-video, image-to-video and combined audio-plus-video generation, with output at 1080p and 4K.
- Native audio. Dialogue, ambience and synchronised effects are generated with the video rather than added afterwards — the feature that most changes whether you still need a separate tool from the AI voice category.
- Ingredients to Video. Multiple reference images pin down characters, objects and style across generations, which is the practical answer to character consistency.
- Frames to Video. Supply a first and last frame and let the model generate the motion between them.
- Extend. Continue from a previous clip to build sequences longer than a single generation.
- Insert and remove. Add elements to an existing scene; the DeepMind page now also lists object removal, which was announced as coming soon at launch.
- Camera and style control. Listed on the model page alongside style matching.
The newer sibling: Gemini Omni Flash
Veo is no longer Google's only video model, and buyers comparing options in late 2026 should know both exist. Gemini Omni Flash was announced at Google I/O on May 19, 2026 as the first model in the Gemini Omni family — a single multimodal model that accepts text, images, audio and video as input and returns video, and that can be edited conversationally in plain English rather than re-prompted from scratch. Public reporting puts developer availability through the Gemini API and AI Studio at June 30, 2026, and describes a 10-second cap per clip.
On pricing, Google's API page lists Gemini Omni Flash on a token basis — $1.50 per million input tokens across text, image, video and audio, with video output at $17.50 per million tokens — and notes this works out to roughly $0.10 per second of video. That lands it alongside Veo 3.1 Fast rather than undercutting it. Crucially, both models remain listed on the pricing page, so Veo has not been deprecated. This is a genuine difference from the situation OpenAI Sora users are in — see Sora alternatives for that migration.
Best-fit use cases
- Prompt-to-clip generation where you want finished audio in the same pass
- Short-form social and ad concepts at Lite or Fast rates, iterating cheaply before committing to a Quality render
- Storyboard and previsualisation work using Frames to Video and Extend
- Developers who need published, predictable list pricing to model a budget before writing code
Limitations to verify
- Credits expire monthly and daily free credits expire daily — an allowance is not a bank balance
- Feature availability varies by tier, platform and region; Google notes phased rollouts and a Japan exception on credit purchases
- Clip length per generation is short by default, so longer pieces mean stitching with Extend or an editor
- Commercial-use rights, likeness rules and watermarking all need checking against your actual output plan
- Per-second billing means cost scales with duration and resolution, not with the number of prompts you try
How it fits against other tools here
Veo is a generator, not an editor or an avatar platform, and that distinction decides most shortlists. If you need a person talking to camera, HeyGen and Synthesia are built for it and Veo is not — the HeyGen vs Synthesia comparison covers that split. If you want one subscription spanning several generative models, Runway lists Veo 3.1 among the third-party models on its own pricing page, which can be cheaper than holding separate accounts. If the job is cutting and captioning footage you already shot, Descript and VEED are the right category. And if you are turning a script or article into faceless social video, Fliki and Pictory are purpose-built for that pipeline. The full list is in the best AI video tools roundup.