Image Generation & Design

Adobe Firefly — Full Introduction (2026, Adobe Inc.) Adobe Firefly is Adobe's commercially-safe generative AI suite​ for images, video, vectors, and audio, launched in beta March 2023 and now sitting across the Creative Cloud ecosystem. Its defining position is training data: Firefly's own models are trained exclusively on licensed Adobe Stock, public-domain works, and openly…

Bannerbear — Template-Based Image/Video/PDF Generation API with AI Layers (BG Removal/Face Detect/AI BG Gen), Workflows, Webhooks, 50+ No-Code Integrations & Official MCP Server (Bootstrap Bali-2018/2020, Active) Bannerbear (sole proprietor / single-member bootstrap, Jon Yongfook Cockle). Origin: Previewmojo (OG-image generator) Nov 2019 → rebranded Bannerbear Jan 2020; REST API + template editor Mar 2020. Legal/base Bali, Indonesia; founder Singapore/Bali. Unfunded, no acquisition, no shutdown.​ ~$1M ARR Sep 2025, 7 staff Jun 2026.

Canva Magic Studio (Canva Pty Ltd, Sydney 2012–2013, 190M+ MAU, 2026 "Creative OS" bundling Affinity + Canva Sheets + Canva Code) = the in-editor AI layer over Canva's drag-drop design surface​ — Magic Write / Magic Design / Dream Lab / Magic Edit-Expand-Grab-Morph-Animate-Switch / Magic Charts-Insights / Conversational AI / Canva Code — not a standalone SKU, not Midjourney, not Figma. 

Clipdrop — Single-Pass AI Image Utility Belt: Remove/Cleanup/Replace Background, Uncrop, Relight, Upscaler, Text Remover, SDXL Text-to-Image, Universal Resizer — Web/iOS/Android/PS Plugin + REST API (Jasper-owned, Acquired-Kept) One uploaded image → pick one atomic action (cutout / inpaint / outpaint / relight / upscale 16x / dedup text / SDXL gen / social resize) → returned in seconds; Stable Diffusion-family models under the hood, Jasper wraps the legacy API into its Business image endpoint.

DALL·E 3 (OpenAI, San Francisco 2015-founded parent; model shipped Oct 2023, API Nov 2023) = the prompt-faithful text-to-image model that lives inside ChatGPT — natural-language prompting + conversational inpainting/outpainting + legible in-image text + C2PA provenance, no standalone app. Not Midjourney (artistic interpretation), not Flux (photoreal king/open-weight), not Ideogram (text-in-image specialist), not Adobe Firefly (license-safe training). Your 39th CPT card, fits ai-productivity-data; pipeline_stage none (spans text-to-image + conversational-edit + api-image-batch). Table-first, Claude Code skeleton. The 2026 trap: DALL·E 3 removed from OpenAI Images API on 2026-05-12​ — new API projects get GPT Image 1, not DALL·E 3; existing integrations grandfathered. Encode that sunset as the headline gotcha.

Fotor = photo-first all-rounder: upload/edit → one-tap AI enhance (unblur/upscale 4K) → bg remover (hair-edge ok) → object eraser → portrait retouch → text-to-image across Flux/GPT Image 2/Nano Banana Pro/Seedream → AI replace/expand/clothes/headshot/avatar → drop into 100K templates (social/poster/collage) → batch edit 50 product shots → export watermark-free on Pro. Fits ai-productivity-data; pipeline_stage none (upload→retouch→generate→template→export). For Etsy/Shopify sellers, social media managers, job-seekers making LinkedIn headshots, and casual users restoring old family photos, it's the cheapest "don't open Photoshop" seat; for top-tier art gen or pro retouch, MJ + Photoshop win.

Ideogram = the "words actually spell right" image model: open ideogram.ai or iOS app → prompt "neon sign saying OPEN 24 HOURS" → pick Ideogram 3.0/4.0 (4.0 open-weight self-hostable) → Magic Prompt expands, bounding box pins text region, JSON for programmatic batches → output 2K PNG with transparency → Magic Fill edit region / Magic Expand uncrop / Remix variants / Character Reference lock a mascot → batch CSV on Pro → same weights via REST API + MCP server for Cursor/Claude Code. Fits ai-productivity-data; pipeline_stage none (prompt→layout→render→edit→batch→ship). For poster/ad/social/logo/packaging/print-on-demand teams where text must be legible, it's the only production-grade seat; for painterly art, editorial photoreal, or SVG brand systems, Midjourney/FLUX/Recraft win.

Khroma = the personalized color-discovery bench, not the production token manager: open khroma.co → rapid-pick 50 colors you're drawn to (≈2 min, re-trainable anytime) → neural net emits an endless scroll of combinations rendered as swatch / gradient / typography pairing / image overlay​ views → each combo shows hex + RGB + WCAG AA/AAA contrast for text-on-bg → search/filter by hue/saturation/value/hex → like-block to steer further → copy hex/CSS into Figma/code by hand. Fits ai-productivity-data; pipeline_stage none (pick 50 → train → scroll → filter → copy hex). For solo designers/brand strategists/UI devs escaping "same 3 colors" at blank-canvas stage, it's the cheapest taste-calibration step; for team token systems + ASE export + contrast simulation, Coolors Pro / Adobe Color / Tokens Studio.

Leonardo.Ai = the asset-production bench, not the hero-image brush: log in (free 150 tokens/day) → pick Lucid Origin (generalist) / Lucid Realism (cinematic+img2video) / Phoenix 1.0 / FLUX Dev-Schnell / or route Veo 3 / Sora 2 / Kling / Seedance → text/img2img/ControlNet pose lock → Realtime Canvas paints sketch→render sub-second → train custom model on 10–50 refs (Essential 10 / Premium 20 / Ultimate 50 slots) for consistent mascot/product face → Motion 2.0 turns stills to lip-synced clips → edit in Canvas (inpaint/outpaint/remove.bg/upscale) → publish private (paid) or public (free) → Teams pool tokens + Starter/Growth roles → Creative Engine API (Apr 2026) replicates pipeline inside proprietary tools, pay-as-you-go credits never expire. Fits ai-productivity-data; pipeline_stage none (prompt→train→render→motion→edit→ship). For game/concept/indie studios, e-com variant shoots, and marketers needing on-brand consistency without Midjourney's Discord tax, it's the workbench; for one-off jaw-drop art, Midjourney still wins aesthetics.

Logo Diffusion = the concept bench for designers, not the one-click founder machine: log in (free 24 cr) → text / upload sketch / upload image → pick 1 of 70+ styles (Line-Art, Monogram, Mascot, Scandinavian, Art Deco, Engraved) → generate 4 variants (2 cr each) → Magic Editor inpaints/outpaints/style-transfers ("make my mark out of marble") → Background Remover (2 cr) → Vectorizer to layered SVG (5–8 cr) → 2×/4× creative upscale (5/20 cr) → Mockup Gen on business cards/packaging → export transparent PNG + SVG, private on paid. Fits ai-productivity-data; pipeline_stage none (prompt/sketch→generate→edit→vectorize→mockup→ship). For freelancers/agencies doing 3–5 client marks/mo who hand off Illustrator-ready SVGs, Pro $49 is the sweet spot; for a non-designer founder wanting logo+card+social in 5 min, Looka/Tailor Brands ship faster.

Looka = the founder brand starter, not the designer replacement: land on looka.com → type business name + pick industry (coffee, SaaS, salon…) → choose styles (modern/minimal/vintage/bold/playful) + colors-liked/disliked + a few icon hints → AI assembles dozens–hundreds of concepts balancing icon+wordmark+spacing → open editor (swap font/color/layout/icon, spacing sliders, history tab) → pick winner → pay $65 once → get PNG/SVG/EPS/PDF, transparent + B/W + color, full commercial ownership, unlimited post-purchase edits​ → flip into Brand Kit ($96/yr) for 300+ templates auto-skinned with your hex/fonts, or Brand Kit + Web ($129/yr) for drag-drop site + custom domain + ecommerce ($299/yr, 100 SKUs, 0 txn fee). Fits ai-productivity-data; pipeline_stage none (quiz→generate→edit→buy→kit→site). For a solo founder launching Tuesday with $65 and no designer, it closes the gap; for a fundable startup needing a defensible trademark-grade mark, hire human + do clearance.

Midjourney = the beauty engine, not the production bench: log into midjourney.com (or /imagine in Discord) → type a rich prompt → V7 returns 4 art-directed variants (8–10s std, Draft Mode ~1s ½ cost) → --cref locks a character across scenes, --sref borrows a style, Omni Reference drags subject/style/structure layers, Personalization learns your rated taste, Moodboards steer direction → upscale (1x/2x), Vary Region (inpainting), Pan/Zoom, Remix → on Pro/Mega animate stills to SD/HD video (Relax video on Pro+) → Standard+ gets unlimited Relax queue (slower, no fast burn), Pro+ gets Stealth (private) + 12 fast concurrency. Fits ai-productivity-data; pipeline_stage none (prompt→variants→refine→upscale→animate). For concept art, editorial, brand mood, album/book cover, game key art — it's still the aesthetic benchmark; for on-brand asset batches with consistent mascot + SVG handoff, Leonardo/Recraft win; for signage with legible text, Ideogram.

Pixlr = the "don't install Photoshop" lane: open pixlr.com/e or /x → load JPG/PNG/WebP/PSD​ → in Pixlr X do one-click Cutout/Retouch/Dispersion/Glitch, in Pixlr E work layers/masks/curves/clone/heal → hit AI Generative Fill​ ("change shirt to leather jacket"), Generative Expand​ (drag canvas, relight edges), Remove BG​ (subject matte in 3s), Magic Eraser / Object Remove, Super Scale​ (≤25MP upscale), Batch Editor​ (100 imgs: BG remove + resize + export) → Premium unlocks Private mode (gens not in community), 1,000 AI cr/mo, 4 concurrent, full 10M+ asset lib → Ultra $24.99 (5k cr, 8 concurrent, priority, unlimited fast image gen fair-use) / Ultra MAX $49.99 (10k cr). Fits ai-productivity-data; pipeline_stage none (upload→retouch→gen-fill→remove→batch→export). For bloggers/thumbnail makers/store owners/school labs on Chromebooks, it's the default; for 16-bit RAW retouch + CMYK print + plugin stack, Photoshop/Affinity still win.

Recraft is the vector-native AI image platform for people who ship design files, not wallpapers. Prompt (or drop an image) and get editable SVG logos/icons/illustrations, raster 1K/2K, in-image legible text, mockups, and brand-locked style sets that hold across dozens of assets. Current engine V4.1​ (Standard / Utility / Pro / Vector variants) reads shorter prompts better, cleans icon geometry, and ships Utility variants tuned for flat product-shot predictability. Studio adds infinite canvas, Agentic Mode (context-aware canvas edits), inpaint/outpaint, background remove, vectorize, creative upscale, and third-party models (Flux, GPT Image, Nano Banana, Seedream) inside the same seat. Export SVG/PNG/JPG/PDF/TIFF/Lottie. It wins where output must survive being opened in Figma/Illustrator and resized to 16px or billboard; it loses on cinematic photoreal (→ MJ/Flux) and doesn't make video natively (Pro tier wraps Kling/Veo/Sora/Seedance via credits).

Remove.bg uses a purpose-trained deep CNN + matting transformer to separate foreground from background and emit a transparent PNG, handling hair strands, fur, glass edges, and semi-transparent fabric better than most generalist editors. It ships as web app, Windows/macOS/Linux desktop (batch), iOS/Android, Photoshop plugin, Figma/GIMP/Unsplash integrations, Zapier/Make nodes, and a RESTful API with Python/Ruby/PHP/JS SDKs (flat $0.20/call). Since the Canva acquisition the same engine powers Canva Pro's in-canvas Background Remover. It is the "cutout brick" — not a studio (Photoroom), not a design suite (Canva), not a pixel editor (Photoshop).

Stable Diffusion is the Linux of text-to-image: a permissively-licensed latent diffusion model that turns prompts (and images) into pictures entirely on your own GPU or inside your own backend. Core lineage runs SD 1.5 → SD 2.1 → SDXL 1.0 (1024²) → SD 3.5 Large/Turbo/Medium (MMDiT, ~1 MP), with sibling models Stable Image Ultra/Core, SDXL Turbo (1-step distill), Stable Video / Fast 3D / Stable Audio in the wider family. You drive it through Automatic1111 WebUI, ComfyUI​ (node graphs), Fooocus​ (Midjourney-like simplicity), or the Stability API / DreamStudio. Because weights are open, you can fine-tune on your brand, lock pose with ControlNet, inject style with LoRA, inpaint/outpaint locally, and ship unlimited images at zero per-call cost. It trades out-of-the-box polish (Midjourney) and text-in-image accuracy (DALL·E 3 / Flux) for total ownership.

Uizard is the "non-designer ships a UI in an afternoon" tool. Start five ways: (1) AI Screenshot-to-Design​ (paste any URL, get a fully editable mock), (2) AI Wireframe​ (text/voice → multi-screen app), (3) AI Theme Generator​ (upload logo → colors/typography → applies across every screen), (4) AI Components​ (prompt a navbar, chart, or flow), (5) hand-drawn sketch scan. It runs a hosted design editor with real component libraries (templates, icons, stock), auto-layout, interactive prototypes with hover/click/scroll, multi-user live edit, and one-click code export​ or Figma handoff. It sits between Canva (visual ease, no code) and Figma (design fidelity, designer-led) and Bolt/Lovable (code-first AI app gen) — weaker visual polish than Figma, stronger "sketch to runnable UI" than both.