Open the model selector in Flow Studio and you’ve got options. Lots of them.
Creative work rarely asks for the same thing twice. One moment you might be sketching out a character or testing the look of a scene. The next, you’re trying to keep that character consistent in motion, edit a shot you already like, add dialogue, or build an environment you can actually explore.
That’s why Flow Studio brings a growing range of image, video, audio, and world generation models into Canvas, alongside the creative control of 3D Editor. Seedance. Veo. Wan. Nano Banana. FLUX. Kling. Gemini. Marble. MiniMax. ElevenLabs. Seed Audio. And within those families, you’ll find even more choices built for different ways of working.
You don’t need to know the ins and outs of every model to get started. You just need to know which one might be useful for what you’re trying to do.
So, here’s the cheat sheet.
If you see labels like Fast, Lite, Prime, Pro, Reference, Draft, or Plus, think of them as signposts. Faster or lighter variants can be useful for exploration (and often use fewer credits), Reference models let you bring existing creative material into the process, and Pro or Plus variants can offer additional capabilities depending on the model family.
Cost is worth keeping in mind, too. Exploring with a lighter model can help you use fewer credits, so you can save heavier models and higher resolutions for the shots you’re ready to commit to. Pricing doesn’t always follow how new or capable a model is, so it’s worth checking before you settle on your workflow.
Image Models
Find your creative look with AI image models in Flow Studio
Before you put your ideas in motion, it’s often helpful to use static images to nail the look.
Image models can help you develop characters, environments, storyboards, lighting ideas, and visual references that can become the starting point for other work in Canvas.
Google | Nano Banana
| Model | What it does best |
| Nano Banana | Quickly generate and edit images while you explore a visual direction. |
| Nano Banana 2 | Balance speed, image quality, consistency, and creative control. |
| Nano Banana Pro | Push precision and control for more complex compositions and detailed edits. |
Nano Banana is useful when you want to move quickly, Nano Banana 2 gives you a strong balance for everyday visual development, and Nano Banana Pro is there when precision and more complex image work matter.
Images are also typically a lighter lift than video, so settling on a look here before you move to video can mean fewer video generations, and fewer credits used.
Video Models
Choose the right AI video model for your project in Flow Studio
This is where the model menu expands.
Different video models can give you different balances of speed, fidelity, reference control, editing, and multimodal generation. Here’s what each family brings to Canvas.
ByteDance | Seedance

| Model | What it does best |
| Seedance 2.0 Fast | Quickly explore motion and video ideas. |
| Seedance 2.0 Reference | Guide video generation with existing creative references. |
| Seedance 2.0 Reference Fast | Explore reference-driven ideas with faster turnaround. |
| Seedance 2.0 Reference – Face Animation | Focus on facial animation and character performance. |
| Seedance 2.0 + 2.5 | Create more ambitious sequences with newer Seedance capabilities. |
| Seedance 2.5 Reference | Combine newer Seedance generation with reference-driven control. |
Reach for Fast when you’re exploring. Use Reference when you already have creative material you want the model to respond to. Face Animation gives you a more specialized tool when the performance is the focus.
Alibaba | Wan
| Model | What it does best |
| Wan 3.0 | Flexible multimodal video generation across different kinds of creative inputs. |
| Wan 3.0 Prime | Get the capabilities of Wan 3.0 with faster generation. |
| Wan 3.0 Reference | Guide video generation using existing creative references. |
| Wan 3.0 Prime Reference | Combine reference-driven generation with faster turnaround. |
Start with Wan 3.0 for general creation, move to Prime when speed matters, and choose a Reference option when you already have content you want to carry into the shot.
Google | Veo
| Model | What it does best |
| Veo 3.1 | Push video fidelity with strong realism, camera control, physics, and native audio. |
| Veo 3.1 Fast | Balance faster generation with high visual quality. |
| Veo 3.1 Lite | Explore ideas efficiently when you want to try more variations. |
Lite is a good place to explore, Fast helps you iterate quickly, and Veo 3.1 is there when fidelity is the priority.
Google | Gemini Omni
| Model | What it does best |
| Gemini Omni 1.1 | Create and refine video across text, images, video, and audio with conversational editing. |
Gemini Omni makes iteration feel a little more like working through revisions: keep this, change that, try another direction.
Kuaishou | Kling
| Model | What it does best |
| Kling 03 Pro Video Edit | Transform or refine existing footage instead of starting over. |
| Kling 03 Pro Reference | Guide video generation with an established visual reference. |
| Kling 4.0 | Generate video with audio from a text prompt or image, with longer clips and more control. |
| Kling 4.0 Reference | Video generation using image, video, character, or object references. |
Kling is particularly useful when you’re not starting from a blank page. Use Video Edit when there’s already something in the shot worth keeping, or Reference when an existing visual direction needs to guide what comes next.
MiniMax
| Model | What it does best |
| MiniMax H3 Max | Guide video with text, images, and visual references. |
| MiniMax H3 | Create high quality, realistic videos with cinematic details from text or image. |
| MiniMax H3 Turbo | Explore video ideas from text or images with faster turnaround. |
Black Forest Labs | FLUX
| Model | What it does best |
| FLUX 3 | Generate video and synchronized sound together, including speech, ambience, and effects. |
FLUX 3 is worth exploring when sound is part of the idea from the beginning rather than something you plan to add later.
Audio Models
Give your characters a voice with AI audio and voice generation models
Picture is only part of a scene. A voice can help define a character, change the feeling of a performance, or shape the timing of a moment before everything else is locked.
This gives you another way to explore a performance while you’re still shaping the scene, instead of waiting until the picture is finished to start thinking about how it sounds.
Google | Gemini
| Model | What it does best |
| Gemini 3.1 TTS Flash | Generate expressive dialogue with control over vocal style, pace, delivery, and multiple speakers. |
ByteDance | Seed Audio
| Model | What it does best |
| Seed Audio 1.0 | Generate dialogue, music, sound effects, and ambience. |
ElevenLabs
| Model | What it does best |
| ElevenLabs v3 | Create voiceovers for your videos. |
| ElevenLabs Turbo v2.5 | Generate voiceovers quickly when you’re drafting, testing timing, or trying lots of takes. |
World Building Models
Build explorable worlds with AI
And then there are models that don’t give you a frame at all.
With World Labs’ Marble models, you can generate an explorable environment and start thinking spatially instead.
World Labs | Marble
| Model | What it does best |
| Marble 1.0 Draft | Quickly explore the shape and feel of a generated environment. |
| Marble 1.0 | Generate a fuller explorable world. |
| Marble 1.1 | Explore newer world-generation capabilities for more developed environments. |
| Marble 1.1 Plus | Push world generation further for more ambitious environments. |
Draft gives you a faster way to explore an environment concept before committing to a fuller world generation. From there, the other Marble models give you options for taking the environment further.
This is also where world generation and 3D Editor start to connect.
A generated image gives you a composition. A generated video gives you a sequence of compositions. A world gives you a 3D workspace, so you can explore the environment, find compositions, stage a scene, and make decisions about where the camera actually belongs.
Build a workflow around the idea, not the model
The fun part is that you don’t have to choose one model and stay there.
Start with Nano Banana 2 to develop a character. Use that image as a reference for Seedance or Wan. Try another interpretation with Veo. Generate dialogue with Gemini 3.1 TTS Flash.
If the story needs a place to live, build an environment with Marble and explore it spatially. Move into 3D Editor when you want more hands-on control over the scene, performance, or camera. Then keep developing the idea in Canvas.
An image can become a reference. A reference can become a video. A voice can change a performance. A generated world can become a location. And an unexpected result from one model can send the whole idea somewhere you hadn’t even thought to imagine.
The models give you possibilities. What you do with them is still up to you.
Frequently asked questions about AI models in Flow Studio
Autodesk Flow Studio brings together a range of AI models for image, video, audio, and world generation within Canvas. These include Nano Banana for image generation and editing; Seedance, Wan, Veo, Kling, and MiniMax for video creation; FLUX for video with synchronized sound; Gemini models for video and audio workflows; ElevenLabs and Seed Audio for voice, music, and sound; and World Labs’ Marble for explorable 3D environments. Each model offers different creative possibilities, so you can choose the right approach for your project.
The right AI model in Flow Studio depends on what you’re trying to create. Start with an image model like Nano Banana when developing a character or visual style. Explore video models like Seedance, Wan, or Veo when working on motion and shots. Reference-based models can help guide generation with existing creative material, while other models support video editing, dialogue, or world generation. The best approach is to start with your creative goal, then experiment with models that fit your needs.
Yes, Flow Studio’s Canvas lets you work with different AI models as your project evolves. For example, you might develop a character concept using Nano Banana, use that image as a reference for a video model like Seedance, and explore another interpretation with Veo. You can also use 3D Editor to assemble and direct 3D scenes. This gives you flexibility to explore different creative directions rather than committing to a single model.
AI image models help you generate and refine static visuals, such as character concepts, environments, storyboards, and visual references. AI video models help you explore motion, camera direction, character performances, and moving sequences. In Flow Studio, these workflows can complement each other: an image created in Canvas can become a visual reference for generating video, helping you develop an idea from an initial concept into a moving shot.
Start with your creative goal rather than choosing a model based on its name or specifications. Use faster or lighter model variants to explore ideas, reference-based models when you have existing visual material to guide the result, and models with more specialized capabilities when your project calls for them. Experiment with different models in Canvas, compare results, and refine your creative direction. When you need more direct control over characters, cameras, or staging, bring the work into 3D Editor.