Google video model · Three documented tiers · Available on Nano Banana Pro

Veo 3.1 AI Video Generator

Turn a written scene, a single still, or a defined pair of first and last frames into a short video through the Veo 3.1 workflow running on this platform.

Try Veo 3.1
01

Create a video with Veo 3.1

Choose text-to-video, image-to-video, or first-and-last-frame mode, pick the Fast or Pro routing chip, and review the point estimate before you submit.

LIVE
4s
4s8s

This model does not support reference video input

Results

Creation history

Generated content will appear here
02 / Overview

Veo 3.1 at a glance

Facts reviewed:

Provider
Google, through Google DeepMind
Official model IDs
veo-3.1-generate-001 for the standard tier and veo-3.1-fast-generate-001 for Fast
Documented tiers
Veo 3.1, Veo 3.1 Fast, and Veo 3.1 Lite
Inputs
Text and images; audio is not an input modality
Official clip length
4, 6, or 8 seconds
Official output resolution
720p, 1080p, and 4K on the standard tier; 720p and 1080p on Fast
Official frame rate
24 FPS, container video/mp4
Provenance
Every output carries a SynthID watermark and C2PA content credentials

What Google documents versus what this workspace exposes

The official model card covers three tiers and a wide set of creative controls. This page runs one provider route with a smaller control surface, so the two boundaries are listed apart rather than blended into a single list.

OFFICIAL

Documented by Google

  • Three tiers on one reference page: veo-3.1-generate-001 and veo-3.1-fast-generate-001 are generally available, while veo-3.1-lite-generate-001 is in preview
  • Clip lengths of 4, 6, or 8 seconds, with reference-image-to-video limited to 8 seconds on the standard tier
  • Up to four output videos per prompt, and a 20 MB ceiling on the image-to-video input image
  • 9:16 and 16:9 aspect ratios, 720p and 1080p input, 24 FPS output, and the MP4 container
  • 4K output on the standard tier; the model card stops Fast and Lite at 1080p
  • Reference images for a scene, a character, an object, or a style, now including audio generation
  • Video extension, first-and-last-frame transitions, and precise camera controls
  • Sound generation with synchronized dialogue, plus SynthID watermarking and C2PA content credentials
  • A single serving region, us-central1, and English as the registered prompt language
LIVE PLATFORM

Exposed on this page

  • Text-to-video, image-to-video, and first-and-last-frame routes
  • A Fast or Pro routing chip, with Fast selected by default
  • A fixed 4, 6, or 8 second selector, defaulting to 4 seconds
  • 720p, 1080p, and 4K on one resolution selector, defaulting to 720p
  • Auto, 16:9, and 9:16 aspect ratios, defaulting to 16:9
  • An audio switch that starts off on this routing path
  • One opening image for image-to-video tasks and two frame images for first-and-last-frame tasks

The official ceiling sits above the control surface here, and one label needs translating. The chip named Pro maps to the standard tier that Google publishes as veo-3.1-generate-001; there is no official tier called Veo 3.1 Pro, and no tier called Veo 3.1 Lite on this page. Reference images, video extension, camera controls, object insertion and removal, and multi-video output are all documented by Google and have no control in this workspace. Resolution is where the two sides are easiest to confuse: the model card puts 4K on the standard tier alone, while this workspace offers one 720p / 1080p / 4K selector that is not restricted by the routing chip, and Google's own price page lists 4K rates for Fast even though the model card does not. This page therefore states the model card reading and makes no claim about Fast at 4K. Two smaller differences: the audio switch combined with Fast follows a preview routing path, and the aspect ratio control is applied to text-to-video tasks, while image routes inherit the shape of the supplied frame.

Model brief

What is Veo 3.1?

Veo 3.1 is the video generation family Google publishes through Google DeepMind. Its reference page documents three tiers in one place: the standard tier served as veo-3.1-generate-001, the faster veo-3.1-fast-generate-001, and a preview tier, veo-3.1-lite-generate-001. All three generate clips of 4, 6, or 8 seconds at 24 frames per second in an MP4 container, accept 9:16 and 16:9 shapes, and return up to four videos from a single prompt. The standard and Fast tiers add 4K output, reference images, and image input; the preview tier does not accept image input and does not use reference images.

Beyond the technical table, the model is built around directing rather than describing. Google documents text-to-video, image-to-video, and first-and-last-frame transitions, plus clip extension that continues from the last second of a shot while holding visual and audio consistency. Reference images let you pin a scene, a character, an object, or a visual style across several generations, and Google's own prompt guide reduces the text side to a five-part formula: cinematography, subject, action, context, and style and ambiance. Every output is marked with a SynthID watermark and carries C2PA content credentials, so the material can be identified as generated.

This page connects that model to the video workspace on this site. The workspace runs one provider route rather than the whole API surface, so the mode selector, the routing chip, the duration and resolution menus, the aspect ratio menu, the audio switch, the uploaded image limits, and the point estimate are platform settings. The official model IDs, tier names, clip lengths, resolutions, frame rate, regional availability, and list prices belong to Google and are quoted here as published facts, not as a description of this integration.

03 / Capabilities

Key Veo 3.1 capabilities

Each entry notes whether the capability is documented by Google, exposed in this workspace, or both.

01

Text to video

Describe the subject, the action, the camera work, the environment, the light, and the sound in one prompt. Available on this page.

02

Image to video

Animate a supplied still and direct motion, camera movement, atmosphere, and audio from the prompt. Google notes stronger prompt adherence and better audio-visual quality here than in the previous generation. Available on this page with one opening image.

03

First and last frame

Supply an opening and a closing image and let the model build a physically plausible transition between them, audio included. Available on this page with two frame images.

04

Native audio and dialogue

Google documents synchronized sound covering multi-person conversation, precisely timed effects, ambience, and music, all directed through the prompt. Exposed here through an audio switch that starts off on this routing path.

05

Reference images for consistency

Google documents reference images for a scene, a character, an object, or a whole visual style, so elements survive across separate generations, and the feature now includes audio. Documented only; this workspace has no reference-image route.

06

Clip extension

Google documents extending a clip from the last second of the shot while preserving visual and audio continuity. Documented only; this workspace has no extension control.

07

Camera control

Google exposes framing and movement as an explicit control and documents a working vocabulary covering dolly, tracking, crane, aerial, slow pan, and point-of-view moves. Documented only; here you direct the camera through prompt language.

08

Provenance marking

A SynthID watermark is embedded in every generated video and C2PA content credentials are attached. This is fixed behaviour rather than a setting, and it applies to everything produced from this page.

04 / Scene ideas

Popular Veo 3.1 use cases

Each entry notes whether the capability is documented by Google, exposed in this workspace, or both.

Product and campaign concepts

Prototype hero shots, reveal moments, and mood films before committing to a production budget, with 4K available on the standard tier for finishing work.

Social video in two shapes

Generate the same idea at 16:9 for widescreen placement and 9:16 for vertical feeds straight from the aspect ratio control.

Motion storyboards and previsualisation

Turn a written scene or an approved key frame into moving previz that a director, a client, or a room full of stakeholders can react to.

Character work that has to survive cuts

Google documents reference images for holding a character, an object, or a style steady across shots; on this workspace that consistency is carried by the prompt and by the supplied frames instead.

Atmosphere, light, and camera studies

Test a camera move, a light change, environmental motion, or a sound cue on its own before it costs a shoot day.

Localised advertising with spoken lines

Because dialogue, effects, and ambience are directed in the prompt, a scripted exchange can be rewritten scene by scene and regenerated without a voice session.

05 / Prompt lab

Veo 3.1 prompt examples

Six briefs built on the five-part formula Google publishes. Load one into the workspace, then change the subject, the action, the camera language, the timing, and the audio.

SCENE 01text to video

Cinematic product reveal

Every element of the official formula in one line: cinematography, subject, action, context, and style, with the light source named as the source of the movement.

View prompt
Cinematic macro shot: slow clockwise orbit around a brushed silver travel bottle standing on dark volcanic rock at dawn. A narrow beam of warm sunlight sweeps across the surface, revealing condensation and true reflections. Volcanic plateau, thin mist, no visible branding. Premium commercial lighting, shallow depth of field, quiet wind and distant seabirds.
SCENE 02image to video

Animate a still image

A motion brief for image-to-video: the uploaded still supplies cinematography and context, so the prompt spends its words on action and stability.

View prompt
Hold the subject, framing, and colour of the uploaded image. Add a gentle forward camera push, natural movement in hair and fabric, slow drifting dust in the backlight, and subtle environmental motion. Keep facial features and identity stable across the full clip. Ambient noise: quiet room tone and distant city traffic.
SCENE 03start end to video

First-to-last-frame transition

A transition brief. Google documents this route as a natural change between two supplied images with audio, so the prompt describes the movement between them rather than either endpoint.

View prompt
One continuous, physically plausible move from the first frame to the last. A crane rises steadily while the scene shifts from blue hour into sunrise. Architecture and perspective stay locked, warm light arrives gradually, and no new foreground objects appear. Ambient noise: birds beginning at daybreak, a distant bell.
SCENE 04text to video

Timestamped multi-shot sequence

Google publishes timestamp prompting as a way to direct several distinct shots inside one generation. Each block names its own shot type and closes with an audio cue.

View prompt
[00:00-00:02] Medium shot from behind a hiker with a canvas pack as she pushes through a curtain of wet ferns onto an open ridge. [00:02-00:04] Reverse shot of her face, wind-burned and bright with surprise, as the valley opens behind her. SFX: ferns snapping back into place. [00:04-00:06] Tracking shot following her along the ridge line, boots finding loose scree, camera staying level with her shoulder. [00:06-00:08] Wide high-angle crane shot pulling up and back, showing the ridge as a thin line inside a huge, half-shadowed valley. SFX: a low string swell entering under the wind.
SCENE 05text to video

Dialogue scene with two speakers

Google instructs writers to place spoken lines inside quotation marks and to name the voice quality. This brief also uses the documented effect and ambience cue formats.

View prompt
Two-shot, medium, inside a narrow repair shop after hours, one hanging bulb and long shadows across a workbench covered in radio parts. An older technician in a worn apron looks up from the bench and says in a dry, tired voice, "It is not the battery. It is never the battery." The customer, half hidden in the doorway, laughs under her breath and answers, "Then stop charging me for batteries." Ambient noise: a fluorescent tube ticking, rain against a roller door.
SCENE 06text to video

Describing an outcome instead of an exclusion

Google's guidance on negative phrasing is to describe the scene you want rather than listing what to remove. This brief states the desired state directly.

View prompt
Wide, slow pan across a level salt flat at first light with an unbroken horizon, wet ground reflecting an empty sky, and a single set of footprints running straight away from the camera into the distance. Bare landscape, open sightlines to the horizon, no structures, no vehicles, no people, no tracks other than the one set. Style and ambiance: quiet, cold, wide-open, documentary stillness.
COMPARE

Veo 3.1, Veo 3.1 Fast, and Veo 3.1 Lite, side by side

Every row below is read from the single official reference page for the Veo 3.1 family, which documents all three tiers in one table. This is the only video model on this site whose tier differences can be quoted from one official page rather than assembled across sources.

Veo 3.1, Veo 3.1 Fast, and Veo 3.1 Lite, side by side
ComparisonVeo 3.1Veo 3.1 FastVeo 3.1 Lite
Official model IDveo-3.1-generate-001veo-3.1-fast-generate-001veo-3.1-lite-generate-001
Launch stageGenerally availableGenerally availablePreview
Release dateNovember 17, 2025November 17, 2025April 2, 2026
Retirement dateNovember 17, 2026 or laterNovember 17, 2026 or laterNot published
Clip length4, 6, or 8 seconds4, 6, or 8 seconds4, 6, or 8 seconds
Reference-image tasksLimited to 8 secondsNo stated limitNo stated limit
Output videos per promptUp to 4Up to 4Up to 4
Image-to-video input ceiling20 MB20 MB20 MB
Aspect ratios9:16 and 16:99:16 and 16:99:16 and 16:9
Input resolution720p and 1080p720p and 1080p720p and 1080p
Output resolution720p, 1080p, and 4K720p and 1080p720p and 1080p
Frame rate and container24 FPS in MP424 FPS in MP424 FPS in MP4
Image inputSupportedSupportedNot supported
Reference imagesSupportedSupportedNot supported
Video extensionSupportedSupportedSupported
Sound generationSupportedSupportedSupported
Prompt languageEnglishEnglishEnglish
Serving regionus-central1us-central1us-central1
LIVE PRICING

What Veo 3.1 costs

Google prices the Veo 3.1 tiers in US dollars per video rather than per second. On the standard tier, a 720p or 1080p clip with synchronized audio lists at $0.40 and the silent version at $0.20; at 4K those become $0.60 and $0.40. Fast lists at $0.10, $0.12, and $0.30 with audio for 720p, 1080p, and 4K, and $0.08, $0.10, and $0.25 without. Lite lists at $0.05 and $0.08 with audio, and $0.03 and $0.05 without, at 720p and 1080p only. A single official reference page produces those numbers, and the same page lists 4K rates for Fast even though the model card stops Fast at 1080p, which is the one place the two official documents disagree. All of these are list prices read on September 12, 2026; Google can change them, and no dated promotion for Veo 3.1 appears on the price page. This site charges platform points, so the estimate beside the generate button is the price that actually applies here.

Open the workspace to see the current estimate
01

Platform points

The workspace quotes points before submission. Points are set by this site and are not the official per-video rate.

02

Resolution and duration

This workspace offers 4, 6, or 8 seconds at 720p, 1080p, or 4K on the current provider route. The duration ceiling matches the official maximum of 8 seconds, and the resolution list is a platform control rather than a restatement of the model card.

06 / Workflow

How to use Veo 3.1

  1. 01

    Choose the input route

    Start from text, upload one opening image, or supply a matched pair of first and last frames. Google documents all three, and each maps to a different part of the URL prompt space.

  2. 02

    Write the prompt in five parts

    Use the formula Google publishes: cinematography, subject, action, context, then style and ambiance. Naming the camera move and the light source usually changes the shot more than adding adjectives does.

  3. 03

    Direct the sound

    Put spoken lines in quotation marks, mark effects with an SFX cue, and set the background with an ambient noise cue. Enable the audio switch before submitting if you want sound on this route.

  4. 04

    Set the platform options and read the estimate

    Pick Fast or Pro, then duration, resolution, and aspect ratio. Longer clips and higher resolutions cost more, and the estimate under the button reflects the combination before anything is spent.

  5. 05

    Review before you publish

    Check continuity across the whole clip, the way motion resolves at the edges of frame, whether objects hold their shape, and whether the dialogue lands where the prompt asked for it. Regenerate a single weak beat rather than the whole sequence.

Review before publishing

Limits, pricing, and responsible use

Generated video can carry unstable faces, hands that change shape between frames, objects that merge or drift, motion that breaks at the frame edges, dialogue that lands late, and audio that does not match the scene. Review the full clip and the complete soundtrack before publishing anything, and treat the output as a draft rather than a record of something that happened.

The official specification and this workspace are two different layers. Google sets the model IDs, the tier names, the clip lengths, the resolutions, the frame rate, the serving region, and the list prices. This site sets the routing chips, the control set, the uploaded image limits, and the point cost. The chip labelled Pro maps to the standard tier rather than to a product Google publishes under that name, and this page makes no claim about Fast at 4K because Google's own model card and price page disagree on it.

Outputs carry a SynthID watermark and C2PA content credentials, so generated material can be identified as generated. That marking is part of the model rather than a setting on this page, and removing or obscuring it would defeat the intended use of the model.

Upload only images you are authorised to use. Do not build deceptive impersonations, misrepresent a real person speaking or acting, infringe privacy or intellectual property, or put generated footage to uses that breach platform policy or applicable law.

Sources

Sources and fact review

Video model capabilities, availability, and prices change, so official Google documentation and platform information are listed separately. Opening the Gemini API guide requires signing in with a Google account. The tier table and the prompt formula on this page come from the official pages below, read on September 12, 2026.

07 / FAQ

Veo 3.1 FAQ

Can I use Veo 3.1 online on this page?
Yes. The workspace above runs the Veo 3.1 family, and you can switch to another video model from the model menu. Submitting a task needs an account and enough platform points.
Can Veo 3.1 turn an image into a video?
Yes. Choose image-to-video and upload an image you are permitted to use. The image becomes the opening visual; the prompt directs the motion, the camera, the atmosphere, and the sound. Google also documents an improved image-to-video path in this generation, with stronger prompt adherence and better audio-visual quality.
Can I control the first and the last frame?
Yes. The first-and-last-frame mode accepts two images and builds a transition between them, and audio can be part of that transition. Google documents this route for all three Veo 3.1 tiers.
Does Veo 3.1 generate sound?
Sound generation is supported on all three official tiers, and Google describes synchronized multi-person dialogue, precisely timed effects, and ambience directed straight from the prompt. Note that the model card lists audio as unsupported for input, which is a different thing: you cannot feed audio in, but the model produces sound.
Which durations and resolutions can I select on this page?
The workspace offers 4, 6, or 8 seconds and 720p, 1080p, or 4K. The duration list matches the official maximum of 8 seconds. The resolution list is a platform control: the official model card puts 4K on the standard tier only, and Google's price page also lists 4K rates for Fast, so this page does not claim either way for Fast.
Is there an official Veo 3.1 Pro tier?
No. Google publishes the family as Veo 3.1, Veo 3.1 Fast, and Veo 3.1 Lite, with the standard tier served as veo-3.1-generate-001. The Pro label you see on this page is a routing chip that maps to that standard tier, not a product name from Google, and you will find Veo 3.1 Pro used on third-party sites because of that shorthand.
What separates Veo 3.1 from Veo 3.1 Fast?
On the official reference page the two share clip lengths, output counts, image ceilings, aspect ratios, input resolutions, frame rate, and capabilities. They differ on output resolution: the standard tier is listed up to 4K while Fast is listed at 720p and 1080p. Fast carries lower list prices at every resolution it does list.
Does Veo 3.1 support reference images for characters and styles?
Google documents reference images for a scene, a character, an object, and an overall visual style, so the same look can survive across separate generations, and the feature now includes audio. The Lite tier is the exception and does not use reference images. This workspace has no reference-image route, so that capability is documented but not available from this page.
Can I extend a clip on this page?
No. Google documents video extension that continues from the last second of a shot while preserving visual and audio continuity, and the model card lists it as supported on all three tiers. There is no extension control in this workspace, so a longer sequence has to be built from separate generations.
How should I structure a Veo 3.1 prompt?
Google publishes a five-part formula: cinematography, subject, action, context, and style and ambiance. The cinematography slot carries the camera move and the shot composition, and Google supplies a working vocabulary for it covering dolly, tracking, crane, aerial, slow pan, and point-of-view moves, plus wide, close-up, extreme close-up, low angle, and two-shot framings. Spoken lines belong in quotation marks, effects are labelled with an SFX cue, and the background is set with an ambient noise cue. For an outcome you want to exclude, describe the scene you do want instead of writing a list of things to avoid.
Are Veo 3.1 videos watermarked?
Yes. Google marks videos made with Veo using SynthID, its watermarking system for detecting AI-generated content, and the model card lists C2PA content credentials as supported. In the same responsible-use section Google notes that outputs pass through safety evaluations and checks for memorised content.
Is there an official Veo 3.1 versus Sora 2 specification comparison?
No side-by-side specification table of that kind is published. Google's DeepMind page reports its own human-preference results and does name competing models in the footnotes, but it does not publish a comparable spec sheet for another vendor's model, and this page will not assemble one from third-party tables, which routinely mix resale pricing into specification rows. What can be compared honestly is the official model information for each model, listed on each model's own page here.
Is Veo 3.1 free on this page?
No unlimited-free claim is made. The workspace shows the current platform point estimate before submission, and the required points move with the routing chip, the duration, the resolution, and the audio switch. The official per-video list prices quoted above are separate from what this site charges.

Related video tools and models

01

Vidu Q3

Compare a 16-second workflow with Pro, Turbo, and an image-to-video Pro Fast route, and a control set that reaches longer clips than the 8-second Veo ceiling.

02

Seedance 2.5

Compare a token-priced model with 4–30 second clips, reference audio and video input, and two extension steps.

03

Wan 2.7

Compare a 2–15 second multi-shot workflow with five aspect ratios and a reference-to-video route this workspace does not offer for Veo.

04

Text to Video

Start from a general video brief and choose between the supported models rather than a single family.

Render your next scene

Render your first Veo 3.1 clip

Return to the workspace, choose an input route, write the prompt in five parts, and review the point estimate before you generate.

Open Veo 3.1 workspace