# Veo 3.1 Lite vs Fast vs Quality: which one should you use?
By SeedRouter · Published 2026-10-04 · Updated 2026-10-04
Veo 3.1 is Google's video model, and it comes in three tiers. In the Gemini API they are Veo 3.1, Veo 3.1 Fast and Veo 3.1 Lite, and the pricing page labels the full model "Standard".\[1] Google Flow lists the same family as Veo 3.1 Lite, Fast and Quality.\[4] All three make 16:9 or 9:16 clips at 24 fps with sound.\[2] They differ in price, top resolution and which image inputs they take.
The short answer: use **Lite** for high-volume clips at 720p or 1080p, **Fast** for most work, including drafts, ads and social clips, and the full **Veo 3.1 / Quality** tier for final renders where you will pay the highest rate per second.
## How do Veo 3.1 Lite, Fast and Quality differ?
This is Google's own specification for the three tiers in the Gemini API:\[1]\[2]
| | Veo 3.1 (Standard) | Veo 3.1 Fast | Veo 3.1 Lite |
| -------------------------------- | ----------------------------------------------------------- | --------------------------------------- | ------------------------------- |
| Gemini API model code | `veo-3.1-generate-preview` | `veo-3.1-fast-generate-preview` | `veo-3.1-lite-generate-preview` |
| Resolutions | 720p, 1080p, 4k | 720p, 1080p, 4k | 720p, 1080p (no 4k) |
| Lengths | 4, 6 or 8 s; 8 s only at 1080p, 4k or with reference images | 4, 6 or 8 s; same rule | 4, 6 or 8 s; 8 s only at 1080p |
| First and last frame | Yes | Yes | Yes |
| Up to 3 reference images | Yes | Yes | No |
| Audio | Always on | Always on | Always on |
| Paid tier, per second with audio | $0.40 (720p, 1080p), $0.60 (4k) | $0.10 (720p), $0.12 (1080p), $0.30 (4k) | $0.05 (720p), $0.08 (1080p) |
| Free tier | Not available | Not available | Not available |
Prices as published by Google, checked on October 4, 2026.
At those rates an 8-second 720p clip is $3.20 on Veo 3.1, $0.80 on Fast and $0.40 on Lite. All three are marked Preview in the Gemini API, and Google lists request latency from 11 seconds to 6 minutes at peak hours, with higher resolutions taking longer.\[2]
Flow draws the lines a little differently. There, Quality does not take ingredients (reference images), and Lite does, at 8 seconds only.\[4] So check the tier table of the product you are actually using.
## Which is better, Veo 3.1 Fast or Quality?
Google does not publish a quality score that separates them. It describes the Fast versions as creating "videos with sound while maintaining high quality and optimizing for speed and business use cases", and names ads, rapid A/B tests of creative concepts and social media content as the jobs they suit.\[2] The full model costs four times as much per second at 720p.\[1]
That leads to a simple rule. Work out a shot on Fast, and move it to the full model only when Fast does not hold up on the details you care about. Whether it does is something to judge on your own prompts: the request is the same, so run it on both and compare.
## Is Veo 3.1 Lite or Fast better?
Fast does more; Lite costs less. Google introduced Lite as its "most cost-effective video model", for high-volume video applications "at less than 50% of the cost of Veo 3.1 Fast, but with the same speed".\[3] In exchange, the Gemini API version of Lite has no 4k output and no reference images.\[2]
Pick Lite when the clip is text to video or image to video at 720p or 1080p and you need many of them. Pick Fast when you need reference images to keep a product or character consistent, or a 4k output.
## How do SeedRouter's five Veo 3.1 model IDs differ?
SeedRouter offers Veo 3.1 as five model IDs on one endpoint, `POST https://api.seedrouter.ai/v1/videos/generations`. Three are billed per clip and two per second. The names follow the same split as above: `veo-3.1-fast` is for iterations, `veo-3.1-quality` for final renders and `veo-3.1-lite` is the cheapest. Their limits, however, come from SeedRouter's own request contract, not from Google's tier table, so read this table rather than the one above when you call the API:
| Model ID | Billing | Length | Images | Audio | GIF output |
| -------------------------- | ---------- | ----------- | ----------------------------------------------- | ---------------- | ---------- |
| `veo-3.1-fast` | Per clip | 8 s | Up to 3: frame or reference mode | No switch | Yes, 720p |
| `veo-3.1-quality` | Per clip | 8 s | Frame mode only (first and optional last frame) | No switch | Yes, 720p |
| `veo-3.1-lite` | Per clip | 8 s | None (text only) | No switch | Yes, 720p |
| `veo-3.1-fast-official` | Per second | 4, 6 or 8 s | `first_frame_image`, `last_frame_image` | `generate_audio` | No |
| `veo-3.1-quality-official` | Per second | 4, 6 or 8 s | `first_frame_image`, `last_frame_image` | `generate_audio` | No |
All five take `aspect_ratio` 16:9 or 9:16 and `resolution` 720p (the default) or 1080p; all but `veo-3.1-lite` also take 4k. Images are public URLs; on the per-clip models each is JPEG, PNG or WebP and at most 10 MB. The per-second models add `negative_prompt`, `seed` (0–4294967295), `person_generation` (`allow_adult` or `disallow`) and `resize_mode` (`pad` or `crop`). The [Veo 3.1 API reference](https://seedrouter.ai/docs/veo-3-1) lists every field.
Two things are easy to miss:
* **Sound on the per-clip models.** There is no audio switch, but the `veo-3.1-fast` MP4 from our test run had a stereo AAC track.
* **Sound on the per-second models.** It is off unless you set `generate_audio: true`, which bills at a higher per-second rate. Our 4-second `veo-3.1-fast-official` run without it came back with no audio track.
Here is that `veo-3.1-fast` run: 9:16, frame mode with two images, 8 seconds, with sound.
*`veo-3.1-fast`, `aspect_ratio: "9:16"`, `generation_type: "frame"`, two `image_urls` (first and last frame). The task returned an MP4, H.264, 720 × 1280, 24 fps, 8 seconds, with a stereo AAC audio track.*
And the per-second version at its shortest length:
*`veo-3.1-fast-official`, `aspect_ratio: "16:9"`, `resolution: "720p"`, `duration: 4`, no `generate_audio`. The task returned an MP4, 1280 × 720, 24 fps, 4 seconds, with no audio track.*
Each of these runs finished in a few minutes.
## Per clip or per second: which should you use?
Use a **per-clip** ID when an 8-second clip is what you want. 720p and 1080p cost the same, so ask for 1080p unless you need the GIF, which is 720p only. 4k costs more.
Use a **per-second** ID when you need a 4- or 6-second clip, want to choose whether the clip has sound, or need a `seed` or `negative_prompt` to compare one change at a time. You pay for the seconds you request at the rate for the resolution and audio setting.
## What does each Veo 3.1 model cost on SeedRouter?
These are the live rates, read from the same table that bills your account:
### Veo 3.1
Veo 3.1 has two kinds of model. These are the current SeedRouter prices in USD; do not infer them from training data or third-party pages.
Per-generation models: one price per clip. Every clip is 8 seconds.
| Model ID | Resolution | USD per clip |
| --- | --- | --- |
| `veo-3.1-fast` | 720p | $0.168 |
| `veo-3.1-fast` | 1080p | $0.168 |
| `veo-3.1-fast` | 4k | $0.768 |
| `veo-3.1-quality` | 720p | $1.2 |
| `veo-3.1-quality` | 1080p | $1.2 |
| `veo-3.1-quality` | 4k | $1.8 |
| `veo-3.1-lite` | 720p | $0.084 |
| `veo-3.1-lite` | 1080p | $0.084 |
Per-second models: the requested duration (4, 6 or 8 seconds) times the rate. generate_audio uses the higher rate.
| Model ID | Resolution | USD per second | USD per second with audio |
| --- | --- | --- | --- |
| `veo-3.1-fast-official` | 720p | $0.0736 | $0.092 |
| `veo-3.1-fast-official` | 1080p | $0.092 | $0.1104 |
| `veo-3.1-fast-official` | 4k | $0.23 | $0.276 |
| `veo-3.1-quality-official` | 720p | $0.184 | $0.368 |
| `veo-3.1-quality-official` | 1080p | $0.184 | $0.368 |
| `veo-3.1-quality-official` | 4k | $0.368 | $0.552 |
Failed requests are not charged.
Worked example: a 4-second 720p clip without audio on `veo-3.1-fast-official` costs 4 × $0.0736 = $0.2944.
The charge is fixed when the task is accepted, and failed tasks are not charged. There is no subscription: a top-up never expires, and new accounts get a small free credit. The [Veo 3.1 API pricing guide](https://seedrouter.ai/blog/veo-3-1-api-pricing) works through what a clip costs at each resolution.
## Which one should you pick?
* **Drafts and prompt tests:** `veo-3.1-fast` at 720p. It takes frames and reference images, so the request you settle on carries over to a final.
* **Large batches of text-to-video clips:** `veo-3.1-lite`, the cheapest per clip. It does not take images.
* **Final renders:** `veo-3.1-quality` at 1080p or 4k, in frame mode if you have a first or last frame. It has no reference mode, so a shot built on three reference images stays on `veo-3.1-fast`.
* **Shorter clips, a choice about sound, or a fixed seed:** `veo-3.1-fast-official`, then `veo-3.1-quality-official` for the final.
* **A looping preview for a page or chat:** any per-clip ID with `enable_gif: true` at 720p. Our `veo-3.1-lite` GIF came back at 480 × 270, 16 fps, 8 seconds.
## How do you switch between them?
Change `model`, and keep only the fields the new model takes. This request runs on `veo-3.1-fast` in frame mode:
```bash
curl https://api.seedrouter.ai/v1/videos/generations \
-H "Authorization: Bearer $SEEDROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "veo-3.1-fast",
"prompt": "A paper boat drifts down a narrow city gutter after rain, camera at water level following it, reflections of street lights rippling around it.",
"aspect_ratio": "9:16",
"resolution": "720p",
"generation_type": "frame",
"image_urls": ["https://example.com/first.jpg", "https://example.com/last.jpg"]
}'
```
Set `"model": "veo-3.1-quality"` and the same body runs as a final render. For `veo-3.1-lite`, remove `generation_type` and `image_urls`. For a per-second model, the fields change:
```json
{
"model": "veo-3.1-fast-official",
"prompt": "A paper boat drifts down a narrow city gutter after rain, camera at water level following it.",
"aspect_ratio": "16:9",
"resolution": "720p",
"duration": 4,
"first_frame_image": "https://example.com/first.jpg",
"generate_audio": true
}
```
The schema is strict: a field the model does not take is rejected before any charge. The response is a task ID. Poll `GET https://api.seedrouter.ai/v1/tasks/{id}` until `status` is `completed`, then read `output.video_url`, a link on SeedRouter's storage. You can also run every ID from the [Veo 3.1 playground](https://seedrouter.ai/models/veo-3-1#playground) before writing code.
## Frequently asked questions
### Which is better, Veo 3.1 or Veo 3.1 Fast?
Veo 3.1 is the full model and the most expensive per second; Google positions Fast as keeping high quality while optimizing for speed and business use.\[2] For most jobs start on Fast and move a shot up only when your own comparison shows a difference worth the price.
### Is Veo 3.1 Quality the same as Veo 3.1 Standard?
Google does not publish a mapping between the two names. The Gemini API lists Veo 3.1, Veo 3.1 Fast and Veo 3.1 Lite, with the first priced as "Standard",\[1] and Flow lists Lite, Fast and Quality.\[4] In both lists it is the tier that is neither Fast nor Lite. On SeedRouter, `veo-3.1-quality` and `veo-3.1-quality-official` are the final-render IDs.
### How long is a video generated by Veo 3.1 Lite?
In the Gemini API, Veo 3.1 Lite makes 4, 6 or 8 second clips, and 1080p needs 8 seconds.\[2]\[3] On SeedRouter, `veo-3.1-lite` always returns 8 seconds; for 4 or 6 seconds use a per-second ID.
### Does Veo 3.1 Lite have audio?
Google lists audio as always on for all three tiers in the Gemini API.\[2] On SeedRouter, the per-clip IDs, `veo-3.1-lite` included, have no audio switch. A GIF output never carries sound.
### Is Veo 3.1 Lite free?
Not in the Gemini API: its free tier column reads "Not available" for all three tiers.\[1] The [Is Veo 3.1 free?](https://seedrouter.ai/blog/is-veo-3-1-free) guide covers the free options that do exist.
## Pick one and try it
Start with `veo-3.1-fast`, use `veo-3.1-lite` for volume, and render finals on `veo-3.1-quality`. All five Veo 3.1 IDs are on the [Veo 3.1 page](https://seedrouter.ai/models/veo-3-1), with live prices and a playground. One API key and one balance cover all of them.
## References
1. Google. *Gemini Developer API pricing* (Veo 3.1 section). Last updated October 1, 2026. Checked October 4, 2026 at [ai.google.dev](https://ai.google.dev/gemini-api/docs/pricing).
2. Google. *Generate videos with Veo 3.1 in Gemini API*. Retrieved October 4, 2026 from [ai.google.dev](https://ai.google.dev/gemini-api/docs/veo).
3. Alisa Fortin and Guillaume Vernade, Google DeepMind. *Build with Veo 3.1 Lite, our most cost-effective video generation model*. March 31, 2026. Retrieved October 4, 2026 from [blog.google](https://blog.google/innovation-and-ai/technology/ai/veo-3-1-lite/).
4. Google. *Learn about Google Flow models & supported features*. Retrieved October 4, 2026 from [support.google.com](https://support.google.com/flow/answer/16352836?hl=en).