New in 0.2.0 Z-Image Turbo, FLUX.1 [schnell] and Qwen Image Edit

Unlimited AI images. Made on your own graphics card.

The Vidoptimus Image Server turns your GPU into an image provider for Vid Optimus. It runs FLUX.2 in a few seconds an image, adds Z-Image Turbo, FLUX.1 [schnell] and Qwen Image Edit on bigger cards, keeps your characters looking the same in every scene, and never sends you a bill. Included with your Vid Optimus licence.

  • No per-image fee
  • 4 models, from 3 seconds an image
  • Up to 4 reference images
  • Private and offline
  • ∞images, no per-image fee
  • 4image models to choose from
  • 3–6 sper image with FLUX.2
  • 4reference images per prompt
  • 6 GBof video memory is enough
Download

Download the Vidoptimus Image Server

Free with your Vid Optimus licence. Install it on the computer with your graphics card.

Windows

Windows 10 and 11, 64-bit

Version 0.2.0 25MB

Download for Windows
  • Needs an active Vid Optimus licence
  • Unlimited images
  • Runs on your own computer

What's new in version 0.2.0

What's new

๐ŸŽจ Z-Image Turbo: photorealism
Very lifelike images, and it's good at rendering readable text inside the picture. Text prompts only (no reference images).
About 14 seconds per 1024ร—1024 image* ยท 10.4 GB download (about 7 GB extra if you already have FLUX.2 Q8)

๐ŸŽฏ FLUX.1 [schnell] 12B: prompt accuracy
A larger 12B model that follows long, detailed prompts closely and handles composition and text very well. Text prompts only.
About 12 seconds per image* ยท 11.7 GB download

๐Ÿง‘โ€๐ŸŽจ Qwen Image Edit 2511: consistent characters and styles
Give it 1 to 3 reference images and it keeps the same face, character or art style in new images. This is useful for a recurring character across a video or a series of thumbnails. It includes a speed-up that brings an edit down to about 47 seconds*.
17.8 GB download

Requirements: the new models need a graphics card with 12 GB of video memory or more. On smaller cards they are marked "Needs a 12 GB+ GPU" in Setup, and FLUX.2 [klein] keeps working as before.

Models · New in 0.2.0

Four image models, one for every kind of picture

FLUX.2 [klein] is the default and runs on almost anything. The other three are optional downloads for graphics cards with 12 GB of video memory or more. All four are Apache 2.0, so the images are yours to use commercially.

FLUX.2 [klein] 4B

The default: fast and all-round

Runs on almost any card from 6 GB. Text prompts, editing and up to 4 reference images, so it is the one Vid Optimus uses for consistent characters out of the box.

3 to 6 s an imageUp to 4 references4.9 or 8.3 GB

Z-Image Turbo

New in 0.2.0: photorealism

Very lifelike images, and good at readable text inside the picture. Text prompts only, no reference images.

About 14 s an imageText only10.4 GB, about 7 GB extra beside FLUX.2 Q8

FLUX.1 [schnell] 12B

New in 0.2.0: prompt accuracy

A larger 12B model that follows long, detailed prompts closely and handles composition and text very well. Text prompts only.

About 12 s an imageText only11.7 GB

Qwen Image Edit 2511

New in 0.2.0: same character, same style

Give it 1 to 3 reference images and it keeps the same face, character or art style in new images: a recurring character across a video or a series of thumbnails. A built-in speed-up brings an edit down to about 47 seconds.

About 47 s an edit1 to 3 references17.8 GB
Consistent characters

The same face in scene 1 and scene 40.

FLUX.2 reads up to 4 reference pictures with every prompt. Turn on consistent characters in Vid Optimus and it sends your cast's faces and your style picture each time, so your hero, your narrator's world and your colours stay the same from the first scene to the last. Since Vid Optimus 2.6, each character in your Story Bible gets a reference portrait first, and Qwen Image Edit keeps a face or an art style from 1 to 3 references.

  • Your recurring characters keep their look
  • One style picture sets the look of the whole video
  • Fast, HD and Full HD sizes, all with references
What you get

Your own image provider, with no meter running

Everything an online image service does for your videos, on hardware you already own.

Unlimited, with no meter running

No credits, no per-image fee and no monthly API bill. Re-roll a scene ten times if you want to.

Consistent characters

Vid Optimus sends your cast's faces and your style picture with every prompt, up to 4 reference pictures per image.

Edit and combine images

Give it a picture and say what to change, or combine up to 4 pictures into a new one.

Picks the best engine for you

It checks your graphics card and memory and marks the right engine and model as Recommended.

Fine on 6 GB cards

A smaller 4-bit FLUX.2 model runs on 6 to 8 GB of video memory, and it uses system memory when video memory runs short. The three bigger models need 12 GB or more.

Four models, one server

FLUX.2 [klein] by default, plus Z-Image Turbo, FLUX.1 [schnell] and Qwen Image Edit as optional downloads. Downloading one never changes your default.

Gives your GPU back

After 10 idle minutes the model unloads, so rendering your video gets the whole graphics card.

A studio of its own

Generate images in your browser without Vid Optimus: sizes, seeds, batches, reference pictures and a gallery.

Share one GPU

Run it on the PC with the big graphics card and use it from your laptop over your home or office network.

How it connects

From download to your first image in 3 steps

The setup page picks the right options for your computer. You only paste two lines into Vid Optimus.

  1. 1 Install

    Install the server

    Run the installer. On Windows it installs just for you, with no admin prompt. The setup page opens in your browser by itself.

  2. 2 Setup

    Pick engine and model

    Click the options marked Recommended, then enter your Vid Optimus licence key. FLUX.2 is a one-time download of about 5 to 9 GB that resumes if it stops. Add the bigger models later from Setup, and press Use this on the one Vid Optimus should use.

  3. 3 Connect

    Connect Vid Optimus

    In Vid Optimus, open Settings, then AI Images, then Vidoptimus Image Server. Paste the Server URL and API key the server window shows and click Test connection.

System requirements

A gaming graphics card is plenty

On an RTX 4060 Ti, a 1024×1024 image takes 5.6 seconds and a 16:9 image 3 seconds. AMD and Intel Arc cards use the Vulkan engine, about a quarter slower. The model loads in a few seconds and stays loaded while you work.

Without a usable graphics card the server still runs on your CPU, but an image then takes minutes, not seconds. Intel integrated graphics can't run the model and use the CPU.

  • WindowsWindows 10 or 11, 64-bit
  • macOSA Mac with Apple Silicon (M1 or newer)
  • Graphics cardNVIDIA, AMD or Intel Arc with 6 GB of video memory or more. 10 GB or more uses the higher-quality FLUX.2 model; Z-Image Turbo, FLUX.1 [schnell] and Qwen Image Edit need 12 GB or more.
  • Disk spaceAbout 6 to 10 GB for FLUX.2 and the engine, plus 10 to 18 GB for each extra model you add
  • LicenceAn active Vid Optimus licence
  • InternetOnly for the one-time download and a licence check. It works offline for up to 7 days at a time.
FAQ

Questions about the Image Server

What it needs, what it costs and how it works with Vid Optimus.

Still have a question? Send us a message and we'll help you set it up. Contact us
What is the Vidoptimus Image Server?

A free companion app for Vid Optimus that makes AI images on your own graphics card with FLUX.2, and since version 0.2.0 also Z-Image Turbo, FLUX.1 [schnell] and Qwen Image Edit. Once it runs, Vid Optimus uses it like any other image provider, except there is no per-image fee and no limit on how many images you make.

Is it really unlimited?

Yes. The images are made by your own graphics card, so there are no credits and nothing is counted. The only limit is how fast your card is: about 3 to 6 seconds an image with FLUX.2 on a modern one, 12 to 14 seconds with the bigger models.

Which model should I use?

FLUX.2 [klein] for speed and for consistent characters in Vid Optimus. Z-Image Turbo for very lifelike pictures and readable text in the image. FLUX.1 [schnell] when your prompts are long and detailed. Qwen Image Edit when the same face, character or art style must carry across images, from 1 to 3 references. Press Use this in Setup to make one the model Vid Optimus uses; in the studio you can pick per image.

Do the new models work on my graphics card?

Z-Image Turbo, FLUX.1 [schnell] and Qwen Image Edit need 12 GB of video memory or more. On smaller cards Setup marks them "Needs a 12 GB+ GPU", and FLUX.2 [klein] keeps working as before.

Can I use consistent characters with every model?

With FLUX.2 [klein] (up to 4 reference images) and Qwen Image Edit (1 to 3). Z-Image Turbo and FLUX.1 [schnell] take text prompts only, so keep one of the other two as your default when consistent characters is on.

How much does it cost?

Nothing extra. It is included with your Vid Optimus licence. It needs an active licence to make images, so it does not work with the free trial alone.

Can I keep the same character in every scene?

Yes. Turn on consistent characters in Vid Optimus and it sends your cast's faces and your style picture with each prompt, up to 4 reference pictures per image, so the same person keeps the same look across the whole video.

Which graphics cards does it work on?

NVIDIA cards (the fastest), AMD and Intel Arc cards, and Apple Silicon Macs. A card with 6 GB of video memory is enough; 10 GB or more uses the higher-quality model at the same speed. Without a usable graphics card it falls back to the CPU, which works but takes minutes per image.

Does it slow down rendering my video?

No. After 10 idle minutes it unloads the model and gives the video memory back, so your export gets the whole graphics card. The model reloads in a few seconds the next time you need an image.

Do my prompts or images go online?

No. After the one-time model download, prompts and images stay on your computer. The server only goes online to check your licence, and that check sends no device ID, so it does not use one of your activations.

Can I use the images commercially?

Yes. All four models, FLUX.2 [klein] 4B, Z-Image Turbo, FLUX.1 [schnell] and Qwen Image Edit 2511, are released under the Apache 2.0 licence, which allows commercial use.

Can I use it without Vid Optimus?

Yes. It has its own studio in your browser, at 127.0.0.1:9810, with prompts, sizes, seeds, batches, reference pictures and a gallery. It also speaks the same API as OpenAI's image API, so other tools can use it with the URL and key it shows.

Windows shows a SmartScreen warning. What do I do?

Windows can show this for a new release before it builds up a reputation. Choose More info, then Run anyway.

Get started

Stop paying per image. Make them on your own GPU.

Included with every Vid Optimus licence. One flat price, unlimited videos, unlimited images.