Models & Actions

Every video comes down to two choices: which model renders it, and which action it performs. Xotic AI has five models — Custom Video (Nova) plus V1–V4 — each with its own actions and strengths. The action you pick determines which model runs.

The models#

Custom Video — Nova (newest)#

The headline image-to-video model. Pick any image, describe the motion in a sentence or two, and the Nova engine animates it with sound — 5, 8, 10 or 15 seconds, in 720p or HD 1080p. There is no action library to pick from: the prompt is the action, so it can do anything from a slow tease to an explicit scene. Clips take about 8 minutes. Best for: getting exactly the scene you have in mind, with audio. See Image-to-video and the prompting tips.

V1 — Standard#

The original, workhorse model with by far the largest action library — hundreds of actions across realistic and anime. Clips are around 7.5 seconds, and you can choose 16, 32, or 64 FPS plus optional upscaling. It's text-to-video (no starting image needed), and it's the fastest, most varied option. Best for: variety and quick results.

V2 — Long clips, with sound#

The higher-quality model that generates longer clips — up to 30 seconds — rendered in 5-second segments at 32 or 64 FPS with 4K upscaling. Its standout feature: V2 videos have audio. It costs more and takes longer, but it's the choice for your best, longest, and most fully-realized scenes. Best for: long, polished video with sound.

V3 — Custom (do anything)#

A freestyle, do-anything action. Instead of a preset motion, you provide a starting image and describe whatever you want — V3 isn't limited to the action library. It's image-to-video, so it needs a starting image. Best for: getting exactly the custom scene you have in mind. See Image-to-video.

V4 — Talking (lip-sync)#

A lip-synced, talking-portrait model. Give it a portrait image and a line of text (up to ~400 characters), and it animates the portrait so she speaks the words on camera, mouth synced to the audio. Also image-to-video. Best for: making her actually say something to you.

The action library#

Actions are the pre-built motions and scenes — each shown as a preview thumbnail so you can see what you'll get. There are hundreds of them (in V1 alone), and the picker keeps them easy to navigate:

• Realistic / Anime tabs — actions are split by style.
• Search — filter the library by keyword.
• Categories — actions span general, positions, foreplay, and solo scenes.

"Requires image" actions#

The V3 (Custom) and V4 (Talking) actions are image-to-video only — they animate a starting image rather than generating from scratch. In the picker they show a "Requires image" overlay: you can see them, but you can't generate until you've provided a starting image. Pick one without an image and the studio nudges you to choose or generate one first. Full detail in Text-to-video & image-to-video.

Which should I use?#

• Want variety and speed → V1.
• Want the longest, highest-quality clip — with sound → V2.
• Want a fully custom scene with sound from an image → Custom Video (Nova).
• Want her to talk to you on camera → V4 (Talking).