Frequently Asked Questions

MuseCat is pay-as-you-go — no monthly subscription, no contract. The Starter pack at ¥4.9 gives you 50 credits to try every model and resolution; credits stay valid for 30 days and you can pay via WeChat QR or in-WeChat JSAPI.

MuseCat deploys behind a Singapore-backed CDN with Tencent Cloud COS for image storage. Overseas models like Gemini are accessed via dedicated proxy — users in mainland China do not need any VPN or extra setup.

Fully supported. MuseCat is a responsive web app — image generation, video, chat all work on mobile browsers. Inside WeChat, users can pay via WeChat Pay and sign in with WeChat OAuth.

Yes — MuseCat ships as a Progressive Web App. On iPhone, tap "Add to Home Screen" in Safari; on Android, tap "Install App" in Chrome. Once installed, you get a native-feel app icon with shortcuts to Image, Video and Chat.

MuseCat supports phone number + SMS code sign-in. Users in mainland China can also scan a WeChat QR or authorize directly inside the WeChat Official Account. No email is required. MuseCat is pay-as-you-go with no subscription — the ¥4.9 Starter pack gives you 50 credits to try every model.

MuseCat integrates leading models across modalities. Image generation: Gemini 3.0 Pro (Nano Banana Pro), Gemini 2.5 Flash, Nano Banana 2, Doubao Seedream 5.0, GPT Image 1.5 / 2, plus custom ComfyUI workflows. AI chat: Gemini family, Doubao, GPT-5.4 / GPT-5.4 mini. Video generation: Seedance 2.0.

In the bottom generation bar on the home page, open the model dropdown and select "Nano Banana Pro" (which is Gemini 3.0 Pro). Enter your prompt and hit Generate. Gemini 3.0 Pro excels at complex scenes, multi-subject composition and accurate text rendering — it costs more credits than baseline models.

Nano Banana Pro is the internal codename for Google's Gemini 3.0 Pro Image — one of the most-watched AI image models of 2025-2026, known for complex multi-subject composition, world knowledge and high-fidelity text rendering. MuseCat integrates it directly; buy a credit pack to start using it, no Google Cloud account required.

Yes. MuseCat ships with OpenAI's newest GPT Image 2 model — pick "GPT Image 2" from the model dropdown in the bottom bar. The previous-generation GPT Image 1.5 is also available for side-by-side comparison or fallback.

GPT Image 2 is OpenAI's next-generation image model — meaningfully better than 1.5 at text rendering, complex scene understanding and long-prompt alignment. It shines for posters, UI mockups and product shots that need crisp typography. GPT Image 1.5 remains a strong cost-performance pick for everyday creation. Both share the same API key in MuseCat — switch any time.

Image generation: GPT Image 1.5 and GPT Image 2. AI chat: GPT-5.4 and GPT-5.4 mini. All OpenAI calls go through a managed proxy, so users in mainland China can use them without VPN.

MuseCat and Midjourney target different needs. MuseCat is multi-model (Google Gemini, ByteDance Doubao, OpenAI etc.) with a Chinese interface, WeChat Pay support, pay-as-you-go credits, plus video generation, AI chat and a template hub. Midjourney is the gold standard for nuanced artistic style. For mainland China users or multi-model comparison workflows, MuseCat is more convenient.

Seedance 2.0 (ByteDance) and Sora 2 (OpenAI) each have strengths. Seedance 2.0 leads on Chinese cultural content, character motion continuity and first/last-frame completion, and is fully open to users in mainland China — already integrated in MuseCat. Sora 2 is stronger for longer cinematic clips but has higher access barriers. MuseCat is evaluating Sora 2 and other overseas models for future integration.

Nano Banana Pro (Gemini 3 Pro Image) outputs native 4K with 94-96% text-rendering accuracy — best for posters, UI mockups and any commercial work that needs precise typography. Seedream 5.0 leads on "web search + deep reasoning + fuzzy-intent editing" at a lower per-image cost — better for time-sensitive content and iterative refinement. MuseCat ships both side by side; switch with one click using the same prompt.

Nano Banana 2 is the fast, cost-effective Gemini image model — ~8-15s per image, ideal for daily social posts, quick drafts and batch jobs. Nano Banana Pro is Gemini 3 Pro Image with "Thinking" reasoning and native 4K rendering, designed for complex compositions, multi-subject scenes and precise Chinese text. Use Banana 2 for everyday work and Pro for hero shots.

In the bottom generation bar, pick a 4K-capable model (Nano Banana Pro, Seedream 5.0 or GPT Image 2), switch the resolution to 4K, type your prompt and hit Generate. Nano Banana Pro renders 4K natively (every pixel from the model); some others upscale 2K → 4K with AI. A 4K image costs roughly 30-50 credits and takes 20-40 seconds.

OpenAI's Sora 2 is not available in mainland China — neither the App Store nor the official API serve China accounts. MuseCat integrates ByteDance's Seedance 2.0 as the domestic alternative: 4-15 second clips at 1080p, text-to-video, first/last-frame completion and reference-image mode — fully accessible inside China with no VPN, pay-as-you-go.

Today's strongest Chinese-text rendering is Nano Banana Pro (Gemini 3 Pro Image) at ~94-96% accuracy, followed by ByteDance's Seedream series and GPT Image 2. For Chinese posters in MuseCat: pick Nano Banana Pro, spell out the exact copy and font style in your prompt (e.g. "kaiti brush calligraphy", "modern sans-serif"), and let text occupy at least one-third of the canvas for best legibility.

Yes. MuseCat is integrated with WeChat Pay — users in China can scan a QR code on the pricing page or pay with one-tap JSAPI inside WeChat. Order status syncs in real time.

Credit cost scales with model and resolution. Roughly: 1K standard ≈ 5-10 credits, 2K HD ≈ 15-25 credits, 4K Ultra HD ≈ 30-50 credits per image. The exact cost shows live in the generation bar before you commit.

Starter pack credits are valid for 30 days; larger packs have longer validity windows (see the Pricing page). Unused credits expire and are cleared automatically — you'll see in-app reminders before expiry. We recommend buying based on actual usage rather than stockpiling.

Used credits are non-refundable. For unused and unexpired credits in special circumstances, please contact support via email. The exact refund rules follow the Terms of Service — we recommend trying the ¥4.9 Starter pack to test all models before larger purchases.

MuseCat supports common aspect ratios (1:1, 4:3, 3:4, 16:9, 9:16 and more) and resolutions from 0.5K up to 4K — covering avatars, social posts, posters, wallpapers and e-commerce product shots.

Usually 10-30 seconds. Specifics depend on model and resolution: Nano Banana 2 takes ~8-15s; Gemini 3.0 Pro and 4K Ultra HD take ~20-40s. Batch jobs and queued runs may take longer.

Through prompts, MuseCat supports virtually every mainstream style: photorealistic, anime, oil/watercolor painting, 3D render, pixel art, vector illustration, film poster, concept design and more. The Template Hub offers curated styles with one-click generation.

Yes. MuseCat integrates Seedance 2.0 and other video models — text-to-video, image-to-video and first/last-frame completion are all supported, with multiple resolution and duration options.

Yes. With GPT Image 1.5 / GPT Image 2, set "Background: Transparent" in the generation bar — the output PNG includes an alpha channel, ready for product shots, stickers and logo cutouts without manual masking.

The Template Hub is MuseCat's curated library of preset prompts paired with the right model — covering posters, avatars, product shots, anime, guofeng, portraits and more. Open any template, upload your reference (or use the preset prompt as-is), and pay-as-you-go to generate a batch of candidates. The editorial team updates templates regularly; trending ones are surfaced on the home page.

Yes. MuseCat is connected to a ComfyUI backend that can run custom node graphs — including ControlNet pose control, IP-Adapter face preservation, LoRA style transfer and 4090-tier cutout. Regular users invoke pre-packaged workflows from the Template Hub; you don't need to build node graphs yourself.

Yes. Nano Banana Pro and Seedream 5.0 both support "fuzzy-intent editing" — upload an image and describe what to change in plain language ("replace the background with a beach", "swap the dress for red", "remove the watermark in the top-left"). The model identifies the region and repaints it locally — no manual masking needed.

Yes. The Batch Generate feature accepts an Excel upload — every row becomes a task, queued automatically. When all complete, download everything as a single ZIP. Ideal for e-commerce sellers and content creators producing assets at scale.

Yes. Drop or paste up to 14 reference images into the generation bar — the AI synthesizes their style, composition and color into new outputs. Gemini, Doubao Seedream and ComfyUI all support image-to-image, perfect for product variants, outfit changes and style transfer.

Fully supported. All integrated image and video models accept mixed Chinese/English prompts. For Chinese aesthetics (guofeng, hanfu, ink painting) Chinese works best; for international styles (cyberpunk, fantasy, photorealism) mixing English with Chinese tends to give the strongest results.

Two recommended approaches: (1) Lightweight — upload the same reference photo (front-facing portrait) every time and write prompts like "same character, [new scene]"; Gemini and Seedream both preserve key facial features. (2) Advanced — use the ComfyUI workflow template with IP-Adapter FaceID for precise identity locking — ideal for comic strips, manga and recurring e-commerce models.

Yes. Typical workflow: upload a flat product shot as reference, then use Gemini 3 Pro or Seedream 5.0 to generate "model wearing" or "in-context scene" shots. The Batch Generate feature accepts an Excel of multiple SKUs, queues them automatically and outputs a single ZIP. Pick GPT Image 2 with transparent background to skip manual masking.

Absolutely. Use Nano Banana Pro (best Chinese text rendering); pick 3:4 or 1:1 for Xiaohongshu, 16:9 for WeChat OA covers, and 9:16 for vertical Douyin/Xiaohongshu covers. Spell out the exact title text, subtitle and style ("minimalist textured", "vintage grain", "Mondo poster") in the prompt — keep text on at least one-third of the canvas. Run a batch of four candidates per prompt to pick the strongest.

Yes. Upload a front-facing photo as reference, then pick a template from the Template Hub — "AI Portrait", "WeChat Avatar", "Q-version Avatar", "Anime Avatar" or "3D Avatar". Templates ship with the right prompt and model; pay-as-you-go and get a batch of variants. For consistent faces across many images, use the ComfyUI workflow with IP-Adapter FaceID.

A practical formula: "quality + subject + style + scene/action + camera/lighting + details" — put the most important words first. Modern models like Nano Banana Pro and Seedream 5.0 understand full sentences better than keyword stacks. Example: "A close-up of a latte with delicate foam patterns, on a wooden table, soft morning light, shallow depth of field, film-grain texture." The Template Hub also offers ready-to-copy prompts.

All supported. Just swap style keywords in the same prompt structure: anime — "Japanese anime, Kyoto Animation palette, painterly cel"; guofeng — "ink wash landscape, hanfu fine-line painting, Dunhuang flying apsaras"; photorealism — "shot on Sony A7R, natural light, 35mm lens, HDR". For Chinese cultural subjects, prefer Chinese prompts; for international styles, mixing English with Chinese works better.

You own the usage rights to anything you generate on MuseCat — personal or commercial. Note that legal ownership of AI-generated content varies by jurisdiction, and you must not infringe on third-party rights or distribute illegal content. See Terms of Service for details.

MuseCat itself does not restrict commercial use of generated images. You must, however, comply with the underlying model providers' policies (OpenAI, Google Gemini, ByteDance Doubao all publish their own terms) and any AI-content regulations in your target region. Before commercial use, double-check the output for third-party likenesses, copyrighted characters or trademarked logos.

Generated images are kept in your private gallery (only you can see them) — you must explicitly mark them "public" before they appear in the public Gallery. Uploaded reference images are stored encrypted on Tencent Cloud COS and used only for the current generation call — never for model training. See the Privacy Policy for full details.

Not recommended — and some underlying models will auto-refuse. Generating or distributing images of identifiable third parties (especially celebrities and public figures) in mainland China involves likeness rights and the "Provisions on the Administration of Deep Synthesis of Internet Information Services", and may be unlawful. MuseCat blocks obviously non-compliant prompts; please follow your local laws and do not infringe on others' rights.