Which image models can I use in CSuite?

The catalog spans the current flagship image models: Flux 2 (Pro, Max, and the local Klein), Nano Banana, Nano Banana Pro, and Nano Banana 2, Seedream 5 and 4.5, Qwen Image 3 and 3 Pro plus the local Qwen Image and Qwen Image 2.1, GPT Image 2 and the 2.5 Sunburst and Flare tiers, Recraft v4.1 and v4.1 Pro, Recraft v4 Vector and Vector Pro for real SVG, Wan 2.7, Gen-4 Image, Grok Imagine Image 2.0, Meta Muse Image, P-Image, and Z Image Turbo, plus P-Image Upscale for enlarging an image you attach. Cloud models run through Runware, OpenRouter and Replicate with your own API keys, and the GPT Image tiers, the Nano Banana family, Grok Imagine Image and FLUX.2 Pro and Max also run direct on your OpenAI, Google, xAI and Black Forest Labs keys; local models run through the GGML runtime on your machine, and on Apple silicon Macs through Ollama as well. Leave the picker on Auto, the default on a fresh install, and CSuite reads the prompt and picks (a logo goes to a vector model, a quick draft to a fast one), or add any model from Runware, OpenRouter or Replicate's own library under Models. Browse the full list with per-model details at csuite.so/models.

Can I generate images locally, fully offline?

Yes. Z Image Turbo, Flux 2 Klein, Qwen Image, and Qwen Image 2.1 run locally through the bundled GGML runtime (built on stable-diffusion.cpp) on Apple silicon, Windows x64, and Linux x64; on Apple silicon Macs, Z Image Turbo and Flux 2 Klein also run through a second bundled Ollama runtime. You pick the steps and resolution per run, a weight download can be paused, resumed, or cancelled, and a model you've just downloaded appears in the picker without a restart. The status line reads Installing runtime, Downloading model, then Loading model the first time. Once the weights are on disk, generation works completely offline: prompts and outputs stay on your machine, and there is no per-image cost.

Can I edit my own photos, or only generated images?

Any image. Drop PNG, JPEG, or WebP files into your project folder (or drag them straight into the sidebar) and they work exactly like generated images: use them as reference inputs for prompt-based revisions, run the one-click AI tools on them, or open them in the editor to crop, resize, adjust, rotate, flip, annotate, convert, and compress.

What can the one-click AI tools do?

Two jobs that don't need a prompt: upscaling (2× or 4×) and background removal. Pick one from the AI menu above the image and the result is saved as a new file beside the original, never overwriting it. Both run on Runware or Replicate, whichever you've connected, and are billed the same way generation is. A model on OpenRouter still generates as usual, but these two tools themselves always run through Runware or Replicate.

Can CSuite help me write a better prompt?

Yes. Type a rough prompt and hit the wand in the composer: your selected text model rewrites it into a detailed image prompt covering subject, style, lighting, and composition, keeping any constraints you specified. The rewrite follows the model you picked and what is attached: with a reference image it becomes an editing instruction that says what to change and what to leave alone, on a vector model it drops photographic language SVG output cannot use, and on a model with its own negative prompt field it keeps exclusions out of the prompt. Your recent prompts also stay as chips above the composer so you can reuse or tweak one, and every generated image remembers the exact prompt and settings that made it, so you can reload them into the composer from the image's info panel.

Which models support reference images?

Most cloud models in the catalog accept reference images, with Nano Banana 2, Seedream 5, Qwen Image 3, GPT Image 2, and Flux 2 among them. Locally, Qwen Image 2.1 edits the images you attach without sending them anywhere. The composer shows an “Add reference” control whenever the selected model supports it, along with that model's limit on how many references you can attach. A few models take other media as input too (Nano Banana 2 accepts reference videos), and the composer grows the matching picker only when the selected model actually accepts it.

What aspect ratios and output formats are available?

It depends on the model: the generation settings only offer what the selected model actually supports. Common ratios are 1:1, 16:9, 9:16, 3:2, and 4:3, with PNG, JPEG, and WebP output, and real SVG from the Recraft v4 Vector tiers. Under Auto the setting can ride along in the prompt: “red car, 16:9” sets the ratio for that run, and the fragment never reaches the model. After the fact, the convert and quality tools change format or compress any image, and the resize tool has one-click presets for the sizes you actually publish at: Open Graph 1200×630, Instagram square and story, and an X header.

Where are my images stored?

In your project folder, as plain files on your own disk. There's no proprietary library or cloud sync, so you can open, back up, or version the folder with any tool you like.

Features · Image

Generate and edit images, with the model you choose.

Q: What does image generation cost?

CSuite doesn't meter or mark up generation. Cloud models are billed by the platform (Runware, OpenRouter or Replicate), or by a vendor like OpenAI, Google or Black Forest Labs when you go direct, to your account, typically a few cents per image, varying by model and resolution. With a concrete model picked, the composer shows its estimated per-image price before you generate, and the built-in analytics show estimated spend per provider from each model's published rates. The exact figure lives on the platform's own dashboard, and OpenRouter reports the real billed cost. Local models are free to run.

One workspace for the full image loop: prompt a frontier model, revise the result with reference images, upscale it or drop the background in a click, then crop, resize, adjust, annotate, and compress, all on files that live in a plain folder on your disk. Cloud models run on your own API keys; local models never leave your machine.

Download the app Browse all models Read the image guide

Choose

Pick the right model per job.

Switch models without switching apps: Flux 2 Pro for a polished product hero, Nano Banana 2 for precise edits, Recraft v4 Vector for a logo that comes back as SVG, or a local Z Image Turbo for free drafts. One API key per platform (Runware, OpenRouter or Replicate) unlocks everything it hosts, a key from a vendor like OpenAI, Google or Black Forest Labs runs that vendor's own image models direct, Auto picks among them from the prompt, and your prompt history, files, and settings stay in the same workspace no matter which model runs.

ProvidersRunwareCloud · key setOpenRouterCloud · key setReplicateCloud · key setGGMLLocal · 3 image modelsOllamaLocal · macOS · 2 image modelsImage modelsKey valid✦ AutoPicked per promptFlux 2 ProBlack Forest Labs · CloudNano Banana 2Google · CloudSeedream 5 ProByteDance · CloudGPT Image 2OpenAI · CloudQwen Image 3Alibaba · CloudRecraft v4 VectorRecraft · Cloud · SVGZ Image TurboAlibaba · LocalOne key per platform unlocks every model it hosts.

Generate

From prompt to picture.

Describe the shot, pick an aspect ratio and format, and generate. The settings panel only offers what the selected model actually supports, so nothing silently fails, and any input without a control of its own (seed, steps, LoRA strength) waits under More settings. On the Recraft vector tiers the result is a real SVG. Leave the model on Auto and “red car, 16:9” sets the ratio from the prompt, for that run alone.The image shown is a real, unretouched Flux 2 Pro generation from CSuite. The exact prompt is in the bar below it, and the same test renders for every model on its catalog page.

Revise

Revise with a prompt, not a redo.

Attach the image you're refining as a reference and describe only what should change (swap a backdrop, restyle a product, fix a detail) while everything else stays put. Any image works as a reference: a previous generation, or a photo you dropped into the project folder. The composer shows how many references the selected model accepts.

One click

Upscale and cut out backgrounds, no prompt required.

Two jobs that shouldn't need a sentence describing them. Pick Upscale 2× or 4× to take a draft to print size, or Remove background to drop a product onto a clean layout. Each runs through Runware or Replicate, whichever you've connected, and saves the result as a new file beside the original, so the source is never overwritten.

Enhance

A rough idea in, a prompt worth running.

Type the gist and press the wand. Your own text model rewrites it into a full image prompt covering subject, style, lighting, and composition. Attach a reference first and it writes an editing instruction instead, naming what to change and what to leave alone.Images are billed per image, so a run that missed still costs you. The rewrite costs a fraction of a cent.

ComposerEnhance prompt with AIYou typeda coffee shop logo, minimal↓Rewritten by your text model, for the model you pickedSent to the modelMinimal coffee shop logo: a stylised coffee bean centred in a circular badge, clean geometric shapes, solid espresso brown and cream fills, uniform stroke weight, no gradients, an arched COFFEE SHOP wordmark along the top edge.Recraft v4 VectorSVG output, so no lens or lighting talk

Edit

Finish it without leaving the app.

A built-in canvas editor covers the last mile that usually means opening a second tool: crop, resize, adjust, rotate, flip, annotate, convert, compress. Every one of these runs locally on your machine (nothing is uploaded) and saves either in place or as a new copy alongside the original.

Crop to any frame

Reframe the 16:9 espresso-tamper hero to a 1:1 crop for a product grid: drag the marquee with a rule-of-thirds overlay, or snap to a preset ratio. Free-form cropping is there when no preset fits.

Resize with presets or exact pixels

Take a tall poster render down to exactly 1080 px for the web: pick a preset from 240p to 1080p and the width follows the source aspect ratio, or switch to Custom for precise dimensions with an optional aspect lock. Publishing presets (Open Graph 1200×630, Instagram square and story, X header) centre-crop to the target frame instead of stretching it.

Adjust the light

Brightness, contrast, and saturation on three sliders, previewed live on the canvas as you drag. Rescue an underexposed photo or give a flat render some snap, then apply, and exactly what you saw is what gets written.

Rotate and flip

Fix an orientation in a click: quarter-turns left or right, or mirror the image horizontally or vertically. Stack them before applying: a rotate plus a flip saves as one edit.

Annotate for feedback

Mark up a render before sending it back: freehand brush, arrows, rectangles, and text callouts in the color and size you pick, like flagging the one highlight that needs brightening. Undo works mark by mark, and the eraser removes annotations without touching the photo underneath.

Convert between formats

Turn a heavy PNG into a WebP for your site, or a JPEG for a client who needs one: same pixels, right container, saved in place or as a sibling copy.

Compress to size

Dial quality against file size with a live readout: a 3.8 MB portrait becomes a 460 KB attachment at 80% quality, small enough for email without visible loss.

Workspace

The rest of the light table.

The small things you reach for constantly: getting a prompt right, knowing what a run will cost, and moving a finished image somewhere else.

Write the prompt with AIType the gist, hit the wand, and your text model expands it into a detailed prompt covering subject, style, lighting, and composition. Attach a reference first and it writes an editing instruction instead.

Reuse what workedRecent prompts sit as chips above the composer. Every generated image also remembers the model and settings behind it, reloadable into the composer in one click.

Know what it costs firstWith a concrete model picked, the composer shows its estimated price per image before you generate, and a live status line tracks the run from queued to saved.

Image details at a glanceOne panel with dimensions, file size, format, and (for anything CSuite made) the model and full prompt that produced it.

Copy straight outCopy any image to your clipboard from its row menu and paste it into Slack, a deck, or an email, with no export step, no reveal-in-Finder detour.

Files that stay yoursImages are ordinary files in your project folder: drag new ones in, filter a long list by name, and open, back up, or version the folder with any tool you like.

Catalog

Available models & providers.

Every image model in the catalog, in one picker. Cloud models run through Runware, OpenRouter and Replicate, or direct from vendors like OpenAI, Google, Black Forest Labs and more, with your own keys; local models run through the GGML runtime on your hardware (and through Ollama on Apple silicon Macs), free and offline.

RunwareCloud · your API key

OpenRouterCloud · your API key

ReplicateCloud · your API key

ByteDanceCloud · your API key

GoogleCloud · your API key

Black Forest LabsCloud · your API key

OpenAICloud · your API key

RunwayCloud · your API key

xAICloud · your API key

KlingCloud · your API key

OllamaLocal · your hardware

GGMLLocal · your hardware