Describe the design and its wording
Tell the agent what you are making, where it will appear and what it should say. Include the exact headline, supporting text, visual style and intended composition.
Create posters, ads and brand visuals with Ideogram on invideo. Build an image from your brief, refine it with Ideogram 4.5, or create an SVG from text or an existing image with Ideogram Vector 1. Give invideo’s agent the words, references and visual direction you want to work with.
Start with a description and choose a frame or let the model decide with Auto. Ideogram 4.5 Generate offers three quality tiers and up to four images per request, each billed separately.

Ideogram 4.5 Edit takes one to five reference images. The first is the source you want to change; any others guide the result. Use a specific instruction to explore another setting, visual treatment or arrangement.

Use a mask to mark the area you want to edit. For large images, choose Auto context to keep the original resolution.

Ideogram works best when it knows what the image should say and where each element goes. The agent writes that structure for you.
Tell the agent what you are making, where it will appear and what it should say. Include the exact headline, supporting text, visual style and intended composition.
Start with a description for a new image. For an image change, attach the source and any references. For SVG output, use Vector 1 Generate with a prompt, or Vectorize with an existing image.
Check the wording, hierarchy, product details and overall composition. Give the agent a specific next instruction, using the version you want to develop as the source.
Bring the subject, exact headline and desired hierarchy into the same brief. Explore a composition with room for the message, then check the words and small details before use.

Compare different headlines, colour directions and focal points. Generate with a portrait, square or landscape frame suited to the placement instead of relying on one crop for every format.

Create logo concepts, icons and illustrations as SVG graphics with Ideogram Vector 1. Start from a written brief, or convert existing artwork with Vectorize and choose the level of detail you want to keep.

Use an image-change workflow with a product source and references for the setting or treatment. Ask for the specific variation you want, then check the label, proportions and other product details.

Use a sketch as the source for Ideogram 4.5 Edit and describe the intended material, lighting or finish. Additional references can help explain the visual direction.

Supply a room or building image and describe a different material palette, furnishing direction or atmosphere. Review the generated proportions and structure as part of choosing the concept.

Choose the output and the kind of input first. Generation, image changes and vector output use different models and controls.
New image from a text prompt. Up to four outputs per request.
Changes guided by images. One source plus up to four references.
Targeted changes. One source; optional mask and references.
Structured composition / source-led variation. 2K; separate text and image workflows.
Speed-focused image exploration. 1K or 2K; text input.
Create a scalable SVG graphic from a text prompt, with up to five optional reference images. Describe the subject, colours and composition, then choose Low, Medium or High quality. When you attach references, the first is the primary source to transform. Use Generate to create a new graphic or give a source image a new direction.
Convert one existing JPEG, PNG or WebP image into an SVG. Choose Simple for fewer, simpler shapes or Detailed to retain more fine detail. Vectorize works from the image without a text prompt, making it the direct route when you already have artwork you want to convert.
Ideogram handles text-heavy, layout-driven images across brand, advertising and production work. With invideo agent underneath it, each team gets the model without writing structured prompts.

Develop posters, graphic concepts and SVG assets with the model suited to the output.

Explore image and headline combinations for announcements, covers and campaigns.

Compare visual directions using a client brief, exact copy and relevant image references.

Develop key art, pitch imagery and visual concepts as still images.

Collaborate in real time with live cursors to show what everyone's working on.
Turn any script or idea into a shot-by-shot plan, then tweak as needed before generating.

Write your script inside invideo, and ask an AI co-writer for help if you'd like.
Picture Premiere Pro with full AI.
Create custom agents to fill specific roles like cinematographer, music designer, and more.
Backed by the firms behind Stripe, Spotify, Flipkart, and ByteDance.
Loading pricing…
The current catalogue includes Ideogram 4.5 Generate, Edit and Precise Edit; Ideogram 4 and Remix; Ideogram 3; P-Image-Ideogram; and Ideogram Vector 1. Their inputs and controls differ.
Yes. Ideogram 4.5 Edit accepts one to five images, with the first treated as the source. Precise Edit takes one source and optional references, while Ideogram 4 Remix takes one source image.
It accepts up to four references alongside the source, or up to three when a mask is attached. The mask must match the source size, with black marking the area to change and white marking the area to keep.
The image models described here return rendered images. Text inside an image is not automatically a separate editable text layer. Use an image-change workflow to request new wording.
Yes. Ideogram Vector 1 Generate creates an SVG from a text prompt and up to five optional reference images. Vectorize converts one existing JPEG, PNG or WebP image into an SVG, with Simple and Detailed options. Both modes are available to all users on invideo.
No. For example, 4.5 Generate has Low, Medium and High quality, while 4.5 Edit also has Very Low. Check the settings and credit estimate for the selected mode.
For Precise Edit, Auto context with a mask is the documented option for retaining full resolution on large images. With context set to None, images above roughly four megapixels may be scaled down.