Veo AI Video Generator

Veo is Google DeepMind’s video model. Veo 3.1 turns a prompt, a still, or a first and last frame into a clip of 4, 6 or 8 seconds, with the sound generated in the same pass. The limits below are from the Gemini API docs.

01 — Capabilities

One finished shot, sound included

Veo 3.1 is specified as a short clip, not a timeline. Length, resolution and references are parameters, and several of them only work together at 8 seconds.

01
Audio is generated with the picture
Dialogue, effects and ambience come out in the same generation. The prompt can carry audio cues, so the sound is not a track you line up afterwards.
02
Only 4, 6 or 8 seconds
Those are the durations Veo 3.1 accepts. 1080p, 4K, reference images and video extension all require the 8-second setting. There is no 30-second mode.
03
First frame, last frame, three references
You can animate one image, interpolate from a first frame to a last frame, or pass up to three reference images. Veo 3.1 Lite drops reference images. That is the steering kit.
04
Up to 4K, in two aspect ratios
The full model lists 720p, 1080p and 4K at 24 fps, in 16:9 or 9:16. 4K is not on Lite. Higher resolutions take longer and cost more, and extending a clip is limited to 720p.

02 — Prompt to output

Prompts that match the clips above

Each prompt describes the picture on its card. “Use this prompt” sends that text to Create. The generator on this page runs Seedance 2.5, not Veo.

01 · Portrait
PromptClose-up of a young man with curly hair in a grey knit sweater, soft window light. He glances away, then smiles. Slow push-in, shallow depth of field.
02 · Product
PromptA white leather sneaker on a pale grey studio backdrop, soft diffused light. The camera orbits slowly. Stitching stays sharp, the look stays minimal.
03 · Street
PromptA slow dolly down a cobblestone lane in an old European village. Climbing roses on stone walls, café tables in morning sun, a few people walking in the distance.
04 · Character
PromptA young woman with shoulder-length brown hair in an olive t-shirt, late-afternoon gold light. She turns her head and looks off. Breeze in her hair, a handheld feel.

03 — Where it fits

Jobs that fit an 8-second ceiling

Veo’s spec is a single shot with sound. Longer stories have to be cut together somewhere else.

A line of dialogue, already timed
Eight seconds is enough for one spoken beat with the mouth and the line generated together. It is not enough for a conversation.
Vertical, and only vertical or wide
16:9 and 9:16 are the documented aspect ratios. Square, 4:3 and 21:9 are not in the Veo 3.1 parameter list.
You already have both stills
First-and-last-frame interpolation is the case Veo is explicit about: two images, and the motion between them, at 8 seconds.
One camera move, not a scene
A push-in or a short orbit fits. A multi-location sequence does not fit inside 8 seconds without dropping the resolution tricks that also want those 8 seconds.

04 — Verdict

What the spec actually guarantees

Strengths

Native audio, with dialogue cues written in the prompt
Interpolation between a first frame and a last frame
720p, 1080p and 4K at 24 fps on Veo 3.1
Up to three reference images on the full model

Trade-offs

Eight seconds is the maximum, and several controls only exist at that length
Only 16:9 and 9:16
Extending a clip drops the output to 720p
Veo is not available on ImgToVid

05 — Side by side

Veo 3.1, Sora 2 and Kling 4.0

Three clips, three ceilings. Sora’s numbers are from OpenAI’s last API guide. The Sora app closed on April 26, 2026, and the Videos API on September 24, 2026.

	Veo 3.1	Sora 2	Kling 4.0
Developer	Google DeepMind	OpenAI	Kuaishou
Single clip	4, 6 or 8 seconds	4–20 seconds while the API was up	3–30 seconds
Picture	720p, 1080p, 4K at 24 fps	Up to 1920×1080 on Sora 2 Pro	720p, 1080p and 4K listed
Sound	Native, with the picture	Synchronized dialogue and effects	Stereo, with lip sync
Steering	First frame, last frame, up to 3 references	Up to 2 saved characters	Up to 15 references, or up to 10 keyframes
On ImgToVid	Coming soon	Coming soon	Coming soon

06 — How it works

Start a clip on ImgToVid

Step 1
Describe one shot
Subject, action, camera. If sound matters, say what is heard in the same sentence. Veo will not be the model that runs it.
Step 2
Use the generator on this page
It runs Seedance 2.5, because Veo is not offered here. The prompt and the settings carry into Create.
Step 3
Generate in Create
Membership and credits are checked there. Coming back to this page does not switch the model over to Veo.

07 — FAQ

Veo, in plain terms

What is Veo?

Veo is Google DeepMind’s video model. The current documented line is Veo 3.1, including a Fast variant and a Lite variant, available through the Gemini API.

How long is a Veo 3.1 clip?

4, 6 or 8 seconds. 1080p, 4K, reference images and video extension require 8 seconds.

Does Veo generate audio?

Yes. Veo 3.1 generates audio natively with the video, and the prompt can include audio cues such as dialogue and ambience.

Which frames can I lock?

An input image, an optional last frame for interpolation, and up to three reference images on Veo 3.1 and Veo 3.1 Fast. Lite does not take reference images. Aspect ratio is 16:9 or 9:16.

Can I generate Veo videos on ImgToVid?

No. The generator on this page runs Seedance 2.5 so you can draft a shot. Veo itself is marked coming soon.

Can I use the results commercially?

Commercial use of anything you generate on ImgToVid depends on your plan. Check Pricing and our Terms before you publish, and make sure prompts and uploads do not infringe someone else’s rights.