⚡ AI Video Gen | Sorceress Wiki | Sorceress

Open AI Video Gen

AI Video Gen is Sorceress’s video creation workspace for making short AI-generated clips. You can generate from text, animate a still image, guide some models with start and end frames, match motion from reference videos, and organize finished clips in a searchable gallery.

The workspace has two generation sources:

Cloud — choose one or more hosted video models and compare results from the same prompt.
Local GPU — available with Pro. Runs through the Sorceress Local AI server and is shown as Pro in the estimate area.

What it does

AI Video Gen turns a prompt and optional media references into videos. Depending on the selected model and mode, you may see controls for:

Text-to-video — describe a scene and generate from scratch.
Image-to-video — add a start frame and animate it.
Start + end frame control — add both a first and last frame when a model supports it.
Motion Match — provide a character image plus one or more motion reference videos; each motion clip creates its own animation.
Multimodal references — attach reference images, videos, and audio, then refer to them in your prompt with labels such as @Image1.
Multi-model generation — select several compatible cloud models and generate one result per model.

Generated videos appear in the center gallery where you can preview, open, download, favorite, reuse settings, retry failures, share to Prompt Lexicon, send to other tools, or delete.

Requirements

You must be signed in to generate videos.
Some cloud models require a start frame and cannot generate from text alone.
Motion Match requires a Character Image and at least one Motion Reference Video.
Local GPU generation requires Pro access and the Sorceress Local AI server.
If you are not signed in, the page can show demo videos, but generation is disabled until you log in.

Interface overview

AI Video Gen is arranged as three main panels on desktop. On mobile, the same model and prompt panels open from the top toolbar.

Left panel: model source and models

Use the left panel to switch between Cloud and Local GPU.

In Cloud mode, each model card shows information such as generation support, estimated output duration, notable capabilities, and special badges. You can select multiple models at once. If a model has adjustable settings, use the gear icon on its card.

The info icon on a model opens a summary of visible capabilities, such as whether it supports text-to-video, image-to-video, start/end frame control, motion matching, native audio options, camera lock options, available resolutions, durations, or aspect ratios.

In Local GPU mode, the panel shows the Local AI Server status, buttons for Installer, Check, and Refresh, and the available local video model/profile choices. Installed models are marked after checking the installer status. If a local profile is not installed, selecting it opens the Installer.

Center: gallery

The center gallery contains your video history and current results. It includes:

Search by prompt text
All and Favorites filters
Desktop grid size control from 1 to 4 columns
Infinite scrolling for older videos
Batch selection for sending multiple completed videos to AutoSprite
Collections panel when viewing Favorites

Hover a completed video to reveal actions. Click a completed video to open the lightbox.

Right panel: prompt and generation controls

The right panel contains:

Generation mode tabs when selected models expose distinct modes
Prompt input
Optional voice input when supported by your browser
Start frame / character image upload
End frame upload when supported
Motion reference video upload in Motion Match mode
Multimodal image, video, and audio reference upload when supported
Estimate area and Generate button
Local GPU save toggle when using Local GPU

Typical cloud workflow

Open AI Video Gen.
Sign in if needed.
In the left panel, choose Cloud.
Select one or more models.
If mode tabs appear, choose the mode you want. A mode is only usable when all selected models support it.
Enter a prompt unless the current mode allows prompt-optional generation.
Add any required media:
- Start Frame for image-to-video or image-only models.
- End Frame if available and you want last-frame control.
- Character Image and Motion Reference Video for Motion Match.
- Multimodal references when the selected mode shows those sections.
Review the estimate shown at the bottom of the right panel.
Click Generate Video or Generate X Videos.
Watch new cards appear at the top of the gallery. Completed cards play on hover and can be opened in the lightbox.

When multiple cloud models are selected, AI Video Gen creates one video per eligible model. In Motion Match, each motion reference clip creates a separate video, so the Generate button shows the number of clips that will be produced.

Prompt input and voice input

The prompt box accepts a natural-language description. Good prompts usually include:

Subject or character
Action or motion
Setting and mood
Camera movement
Lighting
Style or genre

If your browser supports speech recognition, a microphone button appears next to the prompt label.

Click the microphone button.
Speak your prompt.
Click again to stop, or wait for recording to end.
Final recognized speech is appended to the existing prompt.

While voice input is active, the panel shows a Listening... speak now indicator. Voice input availability depends on your browser.

Start frame and end frame

The Start Frame area is used for image-to-video generation. In Motion Match mode, the same area is labeled Character Image.

You can add an image by:

Clicking the drop area and choosing an image file.
Dragging an image file from your computer.
Dragging supported images from Sorceress galleries or tools.
Sending an image into AI Video Gen from compatible Sorceress workflows.

After upload, a thumbnail appears. Hover the thumbnail and click the remove control to clear it.

The End Frame area appears only when the selected model/mode supports it. It is labeled optional in the UI. Use it when you want the video to end on a specific composition or pose. Add or remove it the same way as a start frame.

Motion Match

Motion Match is a motion-control workflow for compatible models. It uses:

A Character Image to define the subject.
One or more Motion Reference Videos to define motion, action, or camera behavior.

To use Motion Match:

Select a cloud model that supports motion matching.
Choose the Motion Match / video-to-video mode if mode tabs are shown.
Add a character image in the Character Image area.
Add one or more motion reference videos.
Enter a prompt if desired. The Generate button does not require a prompt in Motion Match mode.
Click Generate X Videos.

Each motion clip becomes one generated animation. The motion reference area accepts uploaded video files and videos dragged from the gallery. Supported upload prompts in the UI mention common video formats such as MP4 and MOV.

Motion clips show thumbnails, file names, clip numbers, and duration information while being prepared. Very short motion clips are prepared automatically when needed. If you add multiple clips, the panel displays the number of clips and the number of jobs that will be generated.

If you retry a Motion Match result later, the original motion clip may only be available during the current browser session. If the tool says the clip is no longer available, add a motion reference video again and retry; it will reuse the job’s character image when possible.

Multimodal references

Some models expose a Multimodal mode. When active, the right panel may show reference sections for:

Reference Images — up to 9
Reference Videos — up to 3
Reference Audio — up to 3

Uploaded references are labeled in the UI as @Image1, @Image2, @Video1, @Audio1, and so on. Use those labels in your prompt to tell the model how to use each file.

Example prompt pattern:

Use @Image1 as the main character and replicate @Video1 camera movement.

Reference images are useful for character, style, or scene guidance. Reference videos can guide motion or camera behavior. Reference audio can guide music, sound effects, rhythm, or timing when the selected model supports audio references.

Model settings

Cloud model settings vary by model. Open the gear icon on a model card to adjust its available controls. The current source shows these setting types in the UI:

Dropdown options for model-defined choices such as resolution, duration, or aspect ratio.
On/off toggles for model-defined features such as audio or camera-related options when available.
Sliders for numeric settings.

Some selected settings are summarized directly on the model card. The estimate updates when you change settings, switch modes, or add Motion Match clips.

Your selected cloud models, local profile, model settings, Local GPU save preference, and desktop grid size are saved to your account preferences. AI Video Gen always opens on Cloud mode; you can manually switch to Local GPU when needed.

Local GPU mode

Local GPU mode runs video generation through the Sorceress Local AI server and is available with Pro. The estimate area shows Pro and the UI notes that Local GPU is included with Pro.

To use Local GPU:

Choose Local GPU in the left panel.
Confirm you have Pro access. If not, the Generate button offers an upgrade path.
Use Installer if you need to set up the Local AI server or install the local video model.
Click Check to refresh which local models are installed.
Click Refresh to re-check the Local AI server status.
Select an installed local video profile.
Enter a prompt.
For image-to-video, add a start frame. For text-to-video, no start frame is required.
Click Generate Video.

The UI explains that local image and video generation use the same Local AI server. The local video model loads automatically when you generate. Cold loading or switching models can take several minutes.

Local GPU settings

Open the local model gear menu to adjust:

Number of Frames — 49 to 121, in steps of 4.
Frames Per Second — fixed at 24.
Resolution — 720p landscape or 720p portrait.
Seed — leave blank for random output, type a number for repeatability, or click Random.

The local settings panel also shows Local AI server status, with a Refresh server button.

Save generated videos

Local GPU mode includes a Save generated videos toggle:

On — uploads completed local videos to your library so they remain available later.
Off — completed local videos are shown in the current page session only and may disappear when the page refreshes. Nothing is saved.

Gallery actions

Completed gallery cards play on hover. The card label shows the model name and, when detected, the video resolution. Hover a completed video to reveal actions:

Reroll — generate again with the same prompt and settings.
Favorite — mark or unmark the video as a favorite.
Share to Prompt Lexicon — share the prompt, video, model, optional reference image, and an optional comment with the community.
Download — save the video as an MP4.
Reuse — load the prompt, model, settings, and available reference image back into the controls.
Send to... — send the video to AutoSprite or True Pixel.
Queue for AutoSprite — add the video to an AutoSprite queue and open AutoSprite.
Delete — remove the video from your library.

Failed cards show:

Retry — rerun the failed generation with the same settings when possible.
Remove — delete the failed card.

Generating cards may show status text such as local progress messages or waiting messages. After a generation has been running for a while, a Recheck action may appear. Use it if a card appears stuck.

Click a completed video to open it in the lightbox. From the lightbox you can:

Play the video with controls.
Navigate to previous or next completed videos with on-screen arrows.
Use keyboard arrows to navigate.
Press Escape or the close button to close.
Click Use Prompt to reuse the prompt and settings.
Click Copy to copy the prompt.
Click Download to save the video.
Use Send to... for AutoSprite or True Pixel.
When embedded in WizardGenie, drag the video into WizardGenie Explorer.

Favorites, collections, and search

Use the All and Favorites buttons above the gallery to filter results. The search field filters the currently visible gallery by prompt text.

When viewing Favorites, the Collections panel appears below the gallery. You can open a video collection from there; the gallery switches to that collection and shows a Back control to return.

Sending videos to other tools

AI Video Gen can hand completed videos to other Sorceress tools:

AutoSprite — send one video directly, queue a video, or batch-select multiple completed videos and send them together.
True Pixel — send a completed video into the pixel-art workflow.
WizardGenie Explorer — when AI Video Gen is embedded in WizardGenie, completed videos can be dragged into the Explorer.

To batch-send to AutoSprite:

Click Select above the gallery.
Click completed videos to select them, or use Select All.
Click Send to AutoSprite in the action bar.

Queued AutoSprite videos show a Queued badge in the gallery.

The share button opens a Share to Prompt Lexicon modal. It shows a preview of the video, the prompt, the model, and whether a reference image is attached. You can add an optional comment up to 500 characters.

Click Share to Lexicon to publish the generation to the community, or Cancel to close without sharing. After a video is shared, its share button shows a completed state.

Tips and troubleshooting

The Generate button says “Select a Model”

No cloud model is selected. Choose at least one model in the left panel.

The Generate button says “Add Start Frame”

A selected model or mode requires an image. Add a start frame / character image, or deselect models that require image-to-video input.

The Generate button says “Add Motion Video”

You are in Motion Match mode and need at least one motion reference video.

The Generate button says “Enter Prompt”

Most modes require a prompt. Enter a description before generating. Motion Match can be prompt-optional in the UI.

A mode tab is disabled

A mode is disabled when at least one selected model does not support it. Deselect incompatible models to use that mode.

A generation says “Waiting for capacity…”

The service is busy. The card remains active while AI Video Gen waits and retries automatically.

A video has been generating for a long time

Some video models are slow, and local models may spend several minutes loading. If Recheck appears, click it to request an updated status. In Local GPU mode, also click Refresh in the Local AI Server panel to confirm the server is still available.

Local server is offline

Open Installer for setup instructions, start the Local AI server, then click Refresh. The same Local AI server is used for local image and video generation.

Local model is not installed

Click Check in Local GPU mode. Installed models are marked in the list. If a model is not installed, selecting it opens the Installer.

I need to find local model files

The Local GPU panel includes a storage/uninstall help note showing the local Sorceress storage location to paste into File Explorer.

Retry says the motion reference clip is no longer available

Motion reference clips are not saved between sessions. Add a motion reference clip in the Motion Match panel, then retry the failed job.

I want the same local result again

Use the same prompt, settings, and seed. Leaving the seed blank uses a random seed.

FAQ

Can I select multiple models at once?

Yes. In Cloud mode, selecting multiple compatible models generates one video per model. If the selected mode is Motion Match, each motion reference clip generates its own video.

Can I generate without a prompt?

Most modes require a prompt. Motion Match does not require a prompt in the visible Generate button logic, but adding one can still help guide the result.

Can I drag media between tools?

Yes. AI Video Gen accepts dragged images for start and end frames from supported Sorceress tools, and completed gallery videos can be dragged into Motion Match as motion references. In WizardGenie embed mode, completed videos can also be dragged to WizardGenie Explorer.

Are local videos saved automatically?

Only when Save generated videos is enabled in Local GPU mode. If it is disabled, local videos are temporary and may disappear when the page refreshes.

What can I download?

Completed videos can be downloaded as MP4 files from either the gallery card actions or the lightbox.

AI Video Gen