AI image generation

AI image generator built forvideo scenes

GPT Image 2.6 and Nano Banana 2.1, up to 4K. Shorz draws the still every scene of your script needs, then keeps the same face across every scene with reference images.

A prompt generating a set of cinematic scene stills inside Shorz

How to use the ai image generator in three steps

A generic image tool gives you a pretty picture. A video needs forty of them that look like the same world, with the same person in them. Shorz generates stills for scenes, B-roll and avatars inside the app that assembles them, and takes reference photos so scene 12 still looks like scene 1.

Describing an image for a scene in Shorz
1

Start from your script

In a Text-to-Video project, pick AI Images on the Look card and Shorz writes a picture prompt for every scene. Want one image on its own? Add an Image node in Canvas and describe the frame like a photographer brief: subject, setting, lens, light.

Script-driven or typed

A face reference photo keeping the same character across several generated scenes
2

Add a reference for consistency

Add images of your characters, places and style on the References card: a face, a product, a style plate. The model keeps that identity instead of reinventing it. This is how a character survives a 40-scene video.

Same face, every scene

Choosing an image model and style in Shorz
3

Pick a model and a style

GPT Image 2.6 when the image has to carry legible text, Nano Banana 2.1 for 1K to 4K photographic stills. Pick the model on the Look card, then a style on the Style card, or describe your own look.

Up to 4K

Find it in the app under the Text-to-Video chat (Look → AI Images) · the Extra shots card · Avatar and AI Podcast chats · Canvas → Image node.

Why AI Image Generator in Shorz

A generated image with crisp, legible headline text

Text that isn't gibberish

GPT Image 2.6 renders type reliably, which matters the moment an image has to carry a headline: ad frames, quote cards, signs in a scene. Nano Banana 2.1 takes over when you want maximum resolution and photographic detail.

Two models, two jobs

A generated still being used as an avatar inside Shorz

Images that become the video

Every still becomes a scene, an extra shot over a talking head, or the face of an avatar. Shorz animates it, narrates over it and puts captions on top, all without leaving the app.

Feeds every project

Who uses it, and for what

Story scenes: One still per sentence, all in the same visual world.

Story scenes

One still per sentence, all in the same visual world.

Avatar bases: Generate the presenter, then make them talk.

Avatar bases

Generate the presenter, then make them talk.

Ad frames: Product shots and headlines with real text.

Ad frames

Product shots and headlines with real text.

B-roll stills: Fill the gap when stock footage has nothing.

B-roll stills

Fill the gap when stock footage has nothing.

Shorz speaks MCP

Shorz isn't a locked web app. Connect it to Claude, Cursor, Codex or Gemini and your assistant can drive it directly — one message instead of a panel full of clicks. You review before anything is published.

See how MCP automation works
An AI assistant asks Shorz: Generate 12 cinematic 9:16 stills for my script, keeping the same host in every one, and add them as scenes.

You say

“Generate 12 cinematic 9:16 stills for my script, keeping the same host in every one, and add them as scenes.”

Shorz does

  • 1Read the script scene by scene
  • 2Generate a still per scene with a face reference
  • 3Save them to the project library
  • 4Assign each still to its scene

Every tool ships with the same promises

They're not separate products with separate bills. One app, one credit balance, one export that belongs to you.

  • No subscription — prepaid credits that never expire
  • No watermark on anything you export, ever
  • Full commercial rights on what you create
  • A desktop app for Windows and Mac — your files stay on your machine
  • Auto Edit, Clipping and Faceless include 3 free videos (no card needed)
  • Drivable from Claude, Cursor, Codex or Gemini over MCP

AI Image Generator questions

GPT Image 2.6 at low, medium or high quality, Nano Banana 2.1 at 1K, 2K or 4K, and Nano Banana 2 Lite, the fast and cheap tier at 1K only.

Yes. Add reference images of your characters, places and style (a face photo works best) and the model keeps that identity in every generation. It's the same mechanism the Avatar chat uses to make a presenter who looks like a specific person.

16:9 for YouTube and landscape, 1:1 for square feeds, 9:16 for TikTok, Reels and Shorts. They match the project's format, so a generated still fills the frame with no letterboxing.

A few credits per image (1 credit = $0.01), so a few cents. A single image costs less than one second of AI video, which is why AI images are the practical choice for long videos.

Yes. After the render, open Edit, select a scene and pick New picture, or say what to change ("make scene 2 a night scene"). The change comes back as a new version, and older versions stay.

Still have questions? Ask founder