Sign In

AI Video Generator: Kling 3.0, Veo 3.1, Seedance 2.5 and Four More

Video Lab is Spicy Studio’s AI video generator, with three ways in: text-to-video from a written prompt, image-to-video from a still you upload as the first frame (plus an optional last frame for the ending), and reference-to-video, where Seedance takes reference images, a clip and audio that set the look, the movement and the sound.

Seven models share one prompt box, so trying an idea on Seedance 2.5, Kling 3.0 and Veo 3.1 takes a model switch, not a new workflow. Renders can take a few minutes.

How to make an AI video

  1. Choose a starting point

    Write a prompt, or upload a first frame to animate a picture, plus a last frame if the shot should end on a particular image. For reference-to-video, pick a Seedance model and attach images, a clip or audio.

  2. Set the model, length and format

    Seedance 2.5 runs 4 to 30 seconds at up to 1080p, Kling 3.0 makes 5 or 10 seconds and Veo 3.1 always 8. Pick 9:16 for Reels, TikTok and Shorts, 16:9 for YouTube.

  3. Describe the motion

    Say what moves and what the camera does: “slow push-in on a cyclist at dawn, mist over the road”. Kling Omni O3 can split a clip into up to eight prompt segments.

  4. Generate and review

    The credit cost appears next to Generate before you spend anything, and failed renders are refunded. Download the take you like, or use Reuse Input to try a variation.

Video models in the lab

  • Seedance 2.5

    ByteDance’s newest model: text or image to video, up to 10 reference images plus a reference clip and audio, native sound on by default, 4–30 s at up to 1080p. From 10 credits a second at 480p.

  • Seedance 2.0

    The previous Seedance: 5–15 s at 480p or 720p, up to 7 reference images, optional audio and a camera-fixed switch; 35 credits for 5 s at 480p.

  • Kling 3.0

    Start and end frame control, Standard (720p) or Pro (1080p), 5 or 10 seconds, optional sound. 60 credits for 5 s on Standard.

  • Kling Omni O3

    Physics-aware motion, up to eight prompt segments per clip, end-frame control and optional audio; Standard or Premium, 3–15 s, from 14 credits a second.

  • Google Veo 3.1

    Fixed 8-second clips in 16:9 or 9:16, optionally guided by a first and last frame: 80 credits at 720p, 110 at 1080p, 200 at 4K.

  • Spicy UNCENSORED Video and HappyHorse 1.1 (18+)

    Mature-content models for adults, unlocked by a one-time 18+ confirmation. Spicy UNCENSORED Video runs on Wan 3.0 (2–30 s, with sound); HappyHorse 1.1 is image-to-video without audio. Fictional or consenting adults only; identifiable real people only with documented consent.

Every model’s cost, speed and strengths are compared on the AI Models page.

Guides

Related tools

Video Lab FAQ

How many credits does a video cost?

Most models charge per second and Veo 3.1 per clip: five seconds of Seedance 2.5 at 480p cost 50 credits, five seconds of Kling 3.0 Standard 60, and an 8-second Veo 3.1 clip at 720p 80. The 25 free credits on a new account are less than any clip costs, so video needs a plan (from €10 a month for 600 credits) or a one-time pack (from €2 for 90 credits).

Can the video have sound?

Yes. Seedance 2.5 generates native audio by default, while Seedance 2.0, Kling 3.0 and Kling Omni O3 add it as an option; on Kling that raises the rate, for example from 12 to 18 credits a second on Kling 3.0 Standard. To make a face speak your own recording, use Lip Sync.

How long and how sharp can a clip be?

Up to 30 seconds with Seedance 2.5 (4–30 s) or Spicy UNCENSORED Video (2–30 s); Seedance 2.0 and Kling Omni O3 reach 15 seconds, Kling 3.0 makes 5 or 10 and Veo 3.1 always 8. Seedance 2.5 renders 480p to 1080p, Kling 3.0 720p or 1080p, and Veo 3.1 goes up to 4K.

How do I animate one of my photos?

Upload it as the first frame, describe the motion and add a last frame if you want to fix the ending; JPEG, PNG, GIF and WebP files up to 50 MB work. If people are in the picture, use photos of yourself, of fictional or AI-generated characters, or of someone who has agreed to it, never of anyone who hasn’t.

What is reference-to-video?

A Seedance option that steers the clip with reference images (up to 10 on Seedance 2.5, 7 on 2.0), a reference video and reference audio instead of one first frame. With a reference video, Seedance 2.5 bills the generated seconds plus the reference seconds at 6, 13 or 23 credits a second for 480p, 720p or 1080p.

Can I use the videos commercially?

Under the Terms you own what you generate, subject to the licences of the underlying models. You are responsible for the rights and consents behind what you upload and for how you use the result, and every output carries metadata marking it as AI-generated.

Are there 18+ video models?

Yes: Spicy UNCENSORED Video (Wan 3.0) and HappyHorse 1.1, available after a one-time 18+ confirmation. They are for fictional or consenting adults only; anything involving minors, or real people without documented consent, is prohibited.