01 / IMAGINE
Start with “what if.”
Describe it to the agent (Claude Code, Codex, OpenRouter or a local model). It can plan scenes, generate a shot and inspect timeline frames.
ALPHA · LOCAL GENERATION · YOUR DATA
Your ideas,
in motion.
Turn an idea into scenes, images and a finished video, with picture, sound and music generated on your own GPU or a computer on your network. No cloud, no credits, no account.
v0.3.0 Alpha · Windows & Linux · Free & open source
02 / GENERATE
Picture and sound, together.
Every shot arrives with sound effects, dialogue and music. Extend a clip, bridge two clips, or swap a character.
03 / STAY CONSISTENT
Same cast, every shot.
Build a reference library of characters, places and props, and see which scenes use each one.
04 / KEEP
Your work stays yours.
Generate on this computer or a worker on your own network. Projects are plain folders you control.
NEW IN 0.3.0
More ways to create.
More control over your images, your grade and where you generate.
Use another GPU on your network.
Run a Windows or Linux worker and select it in Settings. Generated media comes back to your project.
Now on Linux.
Choose an AppImage, .deb or .rpm package. Local generation needs an NVIDIA GPU with CUDA 13.
Give every shot its look.
Grade with curves, color wheels and eight creative looks. Your adjustments carry through to export.
Build a character sheet.
Generate four views with shared clothing and identity, then open the result in the image editor.
Video and images, one project.
Every project now has both tabs. Turn an image into a scene or reference, and export either format.
Share your setup.
Exchange generator templates as .slop files and export references as reusable RefMods.
HOW IT WORKS
From idea to export.
- 1
Describe
Tell the agent your idea in plain words.
- 2
Plan scenes
It drafts scenes and shots on the generator board.
See the Generator - 3
Add references
Lock in the look of characters, places and props.
See References - 4
Generate
Render shots with sound, speech and music on your GPU or a LAN worker.
See the Generator - 5
Edit & export
Cut, grade and mix on the timeline, then export an MP4.
See the Timeline
SCENE TYPES
More than text to video.
Start from a prompt, a frame, a pose or footage you already have.
First & Last Frame
The default. Guide a shot from an opening frame towards a closing one.
Continue
Continue from a scene’s saved generation state, with adjustable overlap and a choice of beginning or end.
Pose
Borrow movement and timing from a reference video.
Character Replace
Swap a character in existing footage for one from your library.
Extend
Continue an existing clip past its last frame.
Bridge
Generate the connecting segment between two clips.
AnimateExperimental
Drive a repainted frame with the motion of a video.
BEFORE YOU DOWNLOAD
What you’ll need.
Generation needs a capable GPU, either in this computer or in a worker on your network.
- System
- Windows 10 or 11 (x64): installer, MSI or portable ZIP. Linux (x86_64): AppImage, .deb or .rpm. macOS is planned.
- Graphics
- 24 GB+ VRAM recommended for generation. Windows: CUDA 13 for RTX 50, CUDA 12.8 for RTX 30/40, Vulkan for AMD and Intel. Local generation on Linux needs NVIDIA with CUDA 13.
- LAN workers
- Use a GPU on another Windows or Linux computer. Download a worker package and connect in Settings → Workers.
- Model
- MiniMax H3 weights are a separate download under the MiniMax H3 Community License.
- Release
- Alpha software, so expect rough edges. Windows installations and the Linux AppImage support in-app updates. Portable Windows copies update manually from the releases page.
FAQ
Good questions.
Is Slopus free?
Yes. Slopus is free and open source. The MiniMax H3 model weights are a separate download under their own licence.
Does anything leave my computer?
Generation runs on your GPU unless you select a LAN worker, which receives the files it needs and returns results to your project. The AI agent uses your chosen provider; frames you ask it to inspect are sent to that model. Choose a local model to keep agent work on your own machine too.
Which GPUs work?
24 GB or more of VRAM is recommended for generation. On Windows, RTX 50 cards use CUDA 13, RTX 30/40 use CUDA 12.8, and AMD and Intel use Vulkan. Local generation on Linux needs NVIDIA with CUDA 13. You can also generate on a GPU-equipped LAN worker.
Can I bring my own footage and audio?
Yes. Import video and audio onto the timeline, pick still frames from video references, or edit a PNG or JPG in the Image tab. Sound references let you guide a voice using a 2–15 second clip; they require a Ref2VA generator and an image or video reference in the same scene.
What can I export?
MP4 video (H.264, VP9 or AV1, depending on your hardware) at up to 60 fps in 16:9, 9:16, 1:1 or 4:5, and PNG or JPEG for still images.
Is there a macOS or Linux version?
Linux is available now for x86_64 as an AppImage, .deb or .rpm package. The AppImage supports in-app updates. macOS is still planned.
What’s next?
Standalone music, speech and 3D projects are planned.
LESS SETUP. MORE CREATING.
Got an idea? Make it real.
Download Slopus
v0.3.0 Alpha · Windows & Linux · Free & open source