Kimi K3.1 Pro Is Live on Oimi Agent: Top Assistant for the AI Image & Video Canvas
Kimi K3.1 Pro is now available on Oimi Agent, bringing text generation and image understanding to the same AI image and video creation canvas. Use it to draft copy, break a story into shots, analyze reference images, discuss design direction, and turn the discussion into prompts ready for the next step of production.
No membership required. Kimi K3.1 Pro is the first vision-capable chat model on Oimi Agent with no membership gate — open the canvas and start.
Oimi Agent: understand materials, discuss ideas, and execute creation on the canvas
Oimi Agent is the creation assistant inside the canvas. You can describe what you need in plain language, or select existing text and image nodes and let the Agent work from your materials.
From "help me think of a direction" to "add the plan to the canvas," the whole process happens in one workspace:
- Discuss ideas: organize requirements, then propose copy, visual directions, and storyboard plans.
- Understand materials: analyze the text, prompts, and images on the nodes you selected.
- Add text: turn approved copy, scripts, or storyboards into standalone text nodes.
- Generate images: propose image generation from a description or reference images, and execute after you confirm.
You review the plan first, then decide whether to generate images or add nodes. Image edits based on a reference create a new image node — the original stays on the canvas.
The division of labor is explicit: the Agent's chat model handles understanding and organization, while images are generated by a separate image tool. For plain text-to-image, no reference images need to be selected — just describe the scene.
How to use Kimi K3.1 Pro
Open the Agent panel on your Oimi canvas, choose Kimi K3.1 Pro in the chat model menu, and enter your request. When you need existing materials analyzed, select the relevant nodes first.
One model request supports up to 10 images, and for multi-image tasks you can state the role of each image directly in the request.
What Kimi K3.1 Pro does inside the Agent
Product promotion: start from the product photo and discuss poster directions
Select a product image and let the Agent analyze the material before proposing directions:
Analyze the subject, color scheme, and composition of this product photo, and give me three poster directions suitable for social media promotion. Each direction should include the scene, on-image copy, and composition suggestions. Do not generate images yet.
Once a direction is chosen, add background, lighting, aspect ratio, and text requirements, then let the Agent prepare the image prompt and propose generation. You settle the creative discussion first, then move into production.
Storyboarding: turn a story outline into executable shots
Select an existing story or script node and let Kimi K3.1 Pro structure the storyboard:
Based on the selected story outline, organize it into three shots. Each shot should include the scene, character action, shot size, camera design, and an image prompt, all in 16:9. Show me the content first and wait for my confirmation before adding anything to the canvas.
After review, the Agent can turn each shot into its own text node, then prepare the image generation tasks shot by shot. The Agent currently generates up to 3 images per request, so longer storyboard projects can proceed in batches.
Reference analysis: turn a visual feeling into a concrete description
When you like a certain look but struggle to describe it, select the reference images and let the Agent break them down:
Analyze the color palette, lighting, composition, and subject placement of these reference images, summarize the shared visual characteristics, and write a prompt that can be used to create a new image in that style.
For multi-image tasks, you can go further and assign each image a role:
Use the first image for the product's appearance, the second for the background color scheme, and the third for composition. Organize a creation plan first and specify which elements should be kept.
Compared with vaguely asking for "a similar style," this kind of discussion pins the requirement down to concrete frames.
Image editing: be explicit about what stays and what changes
Select the image you want to adjust and state the editing goal:
Keep the product and the overall color scheme, change the background to a naturally lit indoor scene, reduce decorative elements, and leave space for a headline. Give me the edit plan first; generate only after I confirm.
Kimi K3.1 Pro can organize the editing prompt around the image content, and the image tool then produces a new image node — so the original and the new version sit side by side on the canvas.
Copy and prompts: make the words fit the finished image
Once the image is done, you can keep working around it:
Based on the selected image, write three sets of promotional copy. Each set should have a short headline and a short body, in a natural tone without exaggerated claims.
If all you need is a better prompt, just ask:
Keep the subject and style of this prompt, add composition, lighting, and material descriptions, and rewrite it as one clear image-generation prompt.
FAQ
What can Kimi K3.1 Pro do inside Oimi Agent?
It handles understanding and organization: discussing copy and visual direction, analyzing the text and image materials you selected, turning approved plans into text nodes, and preparing image prompts and proposing execution. Images themselves are generated by a separate image tool, which the Agent calls only after you confirm.
Can non-members use Kimi K3.1 Pro?
Yes. Kimi K3.1 Pro has no membership gate on Oimi Agent — open the canvas, choose the model, and start. Membership status is irrelevant.
How do I switch to Kimi K3.1 Pro?
Open the Agent panel on your Oimi canvas, select Kimi K3.1 Pro in the chat model menu, and enter your request. When you want existing materials analyzed, select the relevant nodes on the canvas before asking.
How many images can it analyze at once?
One model request supports up to 10 images. For multi-image tasks, state the role of each image in the request — for example, the first for the product's appearance, the second for the background palette. The more specific the breakdown, the more actionable the plan.
Do I need to select nodes before using it?
Two cases: plain text-to-image needs no selected nodes — just describe the scene. But to analyze existing materials, edit based on a reference image, or write copy around a generated result, select the relevant text or image nodes first so the Agent can work from them.
From material analysis and plan discussion to storyboards and image creation, Kimi K3.1 Pro gives Oimi Agent a text-and-vision option with no membership requirement. Open the AI image and video creation canvas, choose the model, and start from the materials in front of you.