CapCut generates still images from text on the web, on Desktop, and on mobile. Pick the model by what the image has to carry, then download at the size your post needs.
CapCut's AI image generator creates a still image from a text prompt on the web, on CapCut Desktop, and in the mobile app, and the routes differ in where you open them, which models they offer, and how you save the file. On the web, select AI tools > AI design, type your prompt, and click Send.
Say you're promoting a neighborhood 5K run and need a square sign-up image with the race name in the sky. The race name calls for a model whose picker description mentions text rendering, and the square post calls for 1:1; on Desktop and in Video Studio, you set both before you generate.
Where to generate: what each route lets you set and how it saves
On the web, you can start from a prompt in AI design or in Video Studio (formerly AI video maker), and CapCut Desktop and the mobile app each have their own AI image panel. For Video Studio, open it from All tools and set Creation mode to Image, then pick the model, resolution, and aspect ratio before you generate; the next section compares its models.
AI design (web)AI image (CapCut Desktop)AI Image (mobile app)Where to open itAI tools > AI design on the CapCut websiteAI media > AI image in the editor's left panelMedia > AI Image after you tap New projectWhat you set before generatingA prompt, plus an optional reference imageA prompt, Seedream 5.0 or Nano Banana Pro, and an aspect ratioA prompt, a model, and an aspect ratioHow you change the resultMore prompts, or tools that adjust lighting, apply filters, upscale, or enhance detailsMore prompts, or color correction, AI sharpening, upscaling, and background replacementFilters, upscaling, and color adjustmentsHow you save itDownload, or Copy as PNGDownload, after you choose format, quality, and resolutionExport to your device
Desktop's Download panel is where you choose the saved file's format, quality, and resolution, so the walkthrough below runs on Desktop first, then on the web.
Pick the Video Studio model by what the image has to carry
As of November 2026, Video Studio's Image mode lists ten models, with Nano Banana Pro selected by default. Resolutions run from 1K to 4K depending on the model. Every model offers 16:9, 3:2, 4:3, 1:1, 3:4, 2:3, and 9:16, and seven of them add Auto.
ModelWhat it's forResolutionAuto ratioGPT Image 2.5Better image results1K, 2K, 4KYesSeedream 4.3Improved lighting and mood, lifelike portraits2K, 4KYesNano Banana Pro (default)Excellent image quality and advanced text rendering2K, 4KNoNano Banana 2.1Google's latest high-efficiency image model1K, 2K, 4KNoGPT Image 2Superior image quality & text rendering2KYesSeedream 5.0 ProExcellent quality, rich cultural understanding1K, 2KYesNano Banana 2 LiteFast image generation1KNoSeedream 5.0Consistent characters & rich world knowledge2KYesSeedream 4.5Outstanding image consistency and multi-image editing2K, 4KYesSeedream 4.0sBatch input and output with versatile visual styles2K, 4KYes
For lettering that has to read, like the race name, use one of the two models whose descriptions mention text rendering: Nano Banana Pro at 2K or 4K, and GPT Image 2 at 2K only. The three Nano Banana models have no Auto ratio, so with Nano Banana Pro you pick 1:1 yourself for the square sign-up post.
Generating at 4K narrows the list to six models: GPT Image 2.5, Seedream 4.3, Nano Banana Pro, Nano Banana 2.1, Seedream 4.5, and Seedream 4.0s. If 4K is missing from the resolution options, the selected model doesn't offer it; switch to one that lists it, such as Seedream 4.5.
On Desktop, the AI image panel offers two of these models, Seedream 5.0 and Nano Banana Pro.
Generate four options on Desktop, change one thing, then download
- 1
- In CapCut Desktop, enter the editing interface and select AI media > AI image in the left panel. 2
- Type the prompt. An example for the race post: Runners crossing a finish-line banner on a tree-lined street, early-morning golden light, film-photo style, runners in the lower half with open sky above, "Maple Street Run" written across the sky. 3
- Choose Nano Banana Pro, which Video Studio's picker describes as having advanced text rendering, then set the aspect ratio for where the image goes: 1:1 for an Instagram post, 16:9 for YouTube. 4
- Click Send. Desktop returns four results for the prompt; keep the one whose layout is closest to the post. 5
- Decide whether the layout or the look is off. For layout, edit a single phrase and send again, such as "runners in the lower half" changed to "runners in the lower third." For the look, apply one built-in tool: color correction, AI sharpening, upscaling, or background replacement. 6
- Review both versions at the size and layout where the post will appear, and keep the one where the race name reads at that size. 7
- Click Download, choose the format, quality, and resolution, up to 8K, and save the image to your device.
When the still later needs motion, choose Image to video from the same result menu in the AI image panel that holds Download.
On the web, AI design runs the same round in three steps:
- 1
- Select AI tools > AI design, type the race prompt, and click Send. 2
- To change one thing, send a follow-up prompt with one phrase edited, or adjust brightness, contrast, saturation, crop, or filters in the same interface. 3
- Click Download to save the image, or choose Copy as PNG.
In Video Studio's Image mode on the web, you can pick Nano Banana Pro and 1:1 from the model and ratio lists above. In the mobile app, the last step is Export.
Paste the race prompt into the AI image generator
Paste your version of the race prompt into CapCut's AI image generator, and change one phrase per round until the race name reads at post size.
FAQ
Can I generate an image inside a video project?
Yes. In CapCut's online video editor (CapCut Web), open a project and select AI Generate Image in the media panel, then type your prompt. Before it runs, a preview panel shows the credits the action needs and your available balance, and credits are deducted when you click Generate.
Can I start from a photo instead of text alone?
Yes. In AI design, type your prompt, upload a reference image, and click Send. For the race post, upload a photo of a finish line to guide the composition, and use the prompt for what changes: early-morning golden light and "Maple Street Run" across the sky.