Cycling through all six. Tap any point to stop.

Measured on six things

Ideogram AI 4.5 is an image generator and editor built for precise, drift-free refinement. Create native 2K posters, labels, signs, product shots, and more, then paint exactly what you want to change. Its Pixel Lock feature preserves 94-99% of untouched pixels, while maintaining accurate lettering across repeated edits.

Growth

Sharp wedge — mask-locked multi-pass editing with claimed 94-99% pixel preservation — aimed at print and localisation teams whose approved layouts break on re-render. Real need, named buyers. Ceiling is the wrapper: the model isn't theirs.

Screenshots

1 / 1

Ideogram 4.5

About this app

Ideogram 4.5 is a text-to-image and image-editing model released on September 30, 2026, and ideogram-45.com is an independent online studio for running it. I found the headline feature genuinely refreshing: this release is built around multi-turn editing rather than one-shot generation. Instead of re-rendering the whole frame every time you tweak something, you paint a mask over just the area you want to change and describe the edit — the model rewrites only that region while its Pixel Lock feature copies every unpainted pixel straight back from your source. Depending on mask size, a claimed 94-99% of the image comes back bit-for-bit identical, pass after pass, so your tenth revision looks as clean as your first. Lettering remains Ideogram's signature strength — put exact words in quotes and it lays them out like a designer across posters, labels, signage and packaging at native 2K across seven aspect ratios. Two quality modes are offered (standard for quick exploration, High for fi...

Key features

1Mask-based multi-turn editing with Pixel Lock: unpainted pixels copied back from source, claimed 94-99% identity per pass.
2Native 2K output across seven aspect ratios, with lettering that holds exact quoted text on posters, labels and signage.
3Transparent pay-per-image pricing: 42 credits standard / 154 High at $0.01 per credit, failed runs refunded automatically.

Target customer profile

Drawn from the product itself, not from a survey.

Demographic

Graphic & brand designers producing posters, labels and packaging

Pain points

Repeated edits degrade colour, texture and text placement; full re-renders break approved layouts

Primary needs

Precise masked edits, stable typography and multi-pass consistency

Demographic

E-commerce and product marketing teams creating ad and social variants

Pain points

Swapping props, background colours or variant copy means rebuilding the hero shot from scratch

Primary needs

Fast localised edits, multiple aspect ratios and one approved master image

Demographic

Localisation and campaign production agencies

Pain points

Translating signage or updating dates/prices disrupts style, size and positioning of existing designs

Primary needs

Text-only edits that preserve layout, native 2K output and pay-per-image pricing

Scout Review

Written by AI from measured evidence, scored out of 100.

Ideogram 4.5 (the model) is Ideogram's September 2026 release built around multi-turn editing; ideogram-45.com is an independent studio for running it. The site's central claim — paint a mask, describe the edit, and 94-99% of unpainted pixels come back bit-for-bit identical — is demonstrated with labeled percentages across a three-pass sequence rather than asserted. Lettering at native 2K across seven aspect ratios is the second pillar, and the gallery backs it with real prompt examples. Pricing is transparent pay-per-image: 42 credits standard, 154 High, 1 credit = $0.01, with failed runs refunded. The execution is above-average for an indie wrapper: fast, well-designed, honest about its non-affiliation. The gaps are depth (no docs, no compliance posture) and the fundamental one — the differentiator lives in Ideogram's model, not in this site.

Measure by measure

Usability

Prompt box, mask editor, aspect ratios and credit cost are all visible before you commit, and pricing FAQs answer the obvious billing questions. But support is one email address and there is no docs site, changelog or help center in evidence.

Design

Dark studio UI with a working generate/edit panel above the fold and a labeled three-pass pixel-identity demo. Gallery cards each carry a 'Use this prompt' hook. Polished, but branding leans hard on Ideogram's name rather than its own.

Speed

Lighthouse 99 mobile / 100 desktop, LCP 2.1s, CLS 0, TBT 46ms. The site even compresses its own demo images and says so. Genuinely fast, no user complaints about lag found in collected evidence.

Security

Nine of twelve header checks pass — HSTS, CSP, frame protection, Referrer-Policy, Permissions-Policy all set. But HTTPS enforcement flagged off, DMARC missing, and no SOC 2, ISO or pen-test claims anywhere on site.

Accessibility

Lighthouse 97/100 with one failing audit: insufficient contrast between background and foreground colors. Multi-language switcher present. No WCAG statement or a11y page found, so keyboard and screen-reader support are unverified.

Growth

Improvement areas

No documentation, help center or changelog — support is a single email address.
No SOC 2, ISO, pen-test or bug bounty claims; HTTPS enforcement off and DMARC missing in the header probe.
One Lighthouse contrast failure and no WCAG conformance statement on a site aimed at designers.
Differentiator lives in Ideogram's model, not this site — no independent benchmark or third-party validation of the 94-99% claim.

Design & Usability

The design is the strongest part of the pitch. A working generate/edit panel sits above the fold instead of a hero image, and the three-pass demo labels each step with a real pixel-identity number (97.5%, 96.2%, 95.5%) rather than a vague 'no drift' claim. Six use-case cards each ship a 'Use this prompt' button, which is the right instinct for a tool whose value is hard to explain in a headline. Usability is competent but thinner: the studio is browsable without signup, the prompt box caps at 5000 characters with example prompts preloaded, and the pricing page answers the billing questions people actually ask. What's missing is depth — support is a single email address, and there's no docs site, changelog, or help center in the fetched content. For a tool built around a non-obvious workflow (paint a mask, describe the edit, trust Pixel Lock), a real guide would earn more than another gallery card.

Speed & Security

Speed is the cleanest win here. Lighthouse reports 99/100 on mobile (LCP 2.1s, CLS 0, TBT 46ms) and a perfect 100 on desktop, and the site explicitly compresses its own demo images and says so on the page. No lag complaints surfaced in the collected evidence. Security is solid-but-unfinished: nine of twelve header checks pass, including HSTS, CSP, frame protection, Referrer-Policy and Permissions-Policy, but the probe flags HTTPS enforcement as off and DMARC as missing. There is no SOC 2, ISO 27001, pen-test summary or bug bounty anywhere on the site, and the privacy policy depth isn't visible in the fetched content. For an indie front-end to a third-party model that's not disqualifying — but it does mean anyone uploading client work is trusting a brand-new domain with no published compliance posture.

Accessibility & Growth

Accessibility scores 97/100 on Lighthouse with exactly one failing audit: insufficient contrast between background and foreground colors. That's a small, fixable flaw, and it's ironic on a site selling design precision. A language switcher is present, which is more than many indie tools bother with, but there's no WCAG conformance statement or accessibility page, so keyboard and screen-reader support remain unverified rather than confirmed. Growth is the interesting question. The wedge is real and specific: mask-locked multi-pass editing with a claimed 94-99% pixel preservation, aimed at print designers, e-commerce variant teams and localisation agencies whose approved layouts break every time a model re-renders the frame. Midjourney can't do true mask-based editing; Firefly's Generative Fill shifts surrounding pixels; Canva is a beginner tool. That's a genuine gap. The ceiling is structural — ideogram-45.com is an independent wrapper around someone else's model, and the site says so plainly. The domain is three months old with no traction signals in the collected evidence, which is context, not a verdict.

Conclusion

If you're a designer who has watched a poster layout dissolve across five Midjourney re-rolls, this is worth an afternoon. The mask-and-lock workflow is the right answer to a real problem, the pricing is honest, and the site doesn't waste your time. Go in knowing what you're buying: a clean front-end to a model someone else owns, on a domain registered in October 2026, with no published security certifications and a support inbox instead of a help center. Use it for client work if you're comfortable with that; use it for exploration without hesitation. The 94-99% claim is the whole product — test it on your own file before you build a workflow on it.

Suggestions for improvement

Publish a real docs site: mask workflow, Pixel Lock behavior, credit math, and what happens when an edit fails mid-pass.
Add a WCAG conformance statement and fix the one contrast failure — it's a small fix on a site selling design precision.
Turn on HTTPS enforcement and add a DMARC record; both are free and both currently show as gaps in the header probe.
Show a side-by-side against Midjourney and Firefly on the same source image — the differentiator is visual, so prove it visually.

How we score

Alternatives comparison

Named competitors, point by point. Nobody paid to appear here or to be left out.

	Ideogram 4.5	Adobe Firefly	Midjourney	Canva Magic Studio
Mask-based localised editing with pixel preservation	Yes — paint a mask, Pixel Lock copies unpainted pixels back; claimed 94-99% identity per pass	Yes — Generative Fill inpainting, but surrounding pixels can shift across repeated edits	No true mask-based editing with pixel preservation; refinement is re-roll based	Basic AI editing tools; less precise than mask-locked editing for print work
Text rendering inside generated images	Signature strength — exact quoted words laid out at native 2K across seven aspect ratios	Improving but historically weaker at precise typography; better for compositing	Improved text support in recent versions but not a core strength	Strong — real editable text layers via Canva's design editor, not just generated pixels
Pricing model	Pay per image: $19/$49/$99 monthly plans, 42 credits standard / 154 High at $0.01 per credit	Generative credits bundled with Creative Cloud subscriptions	Flat monthly subscription tiers with GPU-hour based usage	Freemium with Pro subscription; AI features gated by plan tier
Ecosystem and workflow integration	Standalone web studio; no integrations listed in site content	Deep Photoshop and Creative Cloud integration	Discord-first with a web app; limited design-tool integration	Full design platform — templates, brand kits, export pipeline built in
Target user and learning curve	Print and brand designers, e-commerce variant teams, localisation agencies; mask workflow requires learning	Creative Cloud users; familiar if you already know Photoshop	Prompt-craft hobbyists and artists; steep prompt learning curve, no mask precision	Beginners and social marketers; lowest learning curve of the four

Adobe Firefly

Adobe's generative image suite with inpainting (Generative Fill) and tight Photoshop integration — strong for teams already inside Creative Cloud, though repeated edits can shift surrounding pixels.

Midjourney

A popular text-to-image generator known for striking visuals, but it lacks true mask-based editing with pixel preservation, making multi-pass refinement more of a re-roll.

Canva Magic Studio

Beginner-friendly design platform with AI editing tools — great for quick social graphics, but less precise than mask-locked, drift-free editing for professional print work.

Comparing options? See Ideogram 4.5 alternatives, scored side by side

View all Design & Creativity