Ideogram V4.5

• 🎨 Inpainting and editing variant of Ideogram's V4.5 image model
• 🏢 Built by Ideogram, founded 2022 and known for text rendering
• 📅 Released 30 September 2026, alongside the V4.5 generation model
• 🖌️ Takes an input image plus mask and prompt for region edits
• 🔤 Family focus is typography, layout and design-grade output
• 🧩 Successor line to Ideogram V4, released June 2026
• 🌐 Available through Ideogram's hosted platform and API tooling
• 📚 Detailed V4.5 technical documentation not publicly published at time of writing

Ideogram

Ideogram is a text-to-image AI company founded in 2022 by Mohammad Norouzi, William Chan, Chitwan Saharia, and Jonathan Ho. The team set out to build a generative image model with a particular strength that many rivals struggled with: rendering legible,…

Ideogram V4.5 (edit) is the inpainting-oriented endpoint of Ideogram's V4.5 release, paired with the text-to-image sibling Ideogram V4.5. Rather than generating from a blank canvas, it takes an existing image together with a mask and a prompt, regenerating only the selected region so the rest of the composition stays intact — the typical workflow for fixing typography, swapping objects, or extending a layout.

Ideogram, founded in 2022 by Mohammad Norouzi, William Chan, Chitwan Saharia and Jonathan Ho, has built its models around text rendering inside images, and that emphasis carries through the family. The preceding generation, Ideogram V4, was released in June 2026 as a 9.3B single-stream Diffusion Transformer trained from scratch with a vision-language text encoder, structured JSON prompting, explicit bounding-box layout and color-palette controls, and native 2K output; it was also the company's first open-weight foundation model, with fp8 and nf4 checkpoints and inference code published publicly.

V4.5 is the next step in that line, released 30 September 2026. Ideogram has not published a technical report or model card for V4.5 at the time of writing, so architecture details, weight availability, and measured quality changes relative to V4 are not documented from primary sources. For editing workflows, treat it as the current-generation counterpart to the V4 editing stack and verify behavior against your own prompts and masks.