Step-by-Step Guide

How to Use Ernie Image

From account creation to downloaded image in under three minutes. This guide covers every step — registration, prompt writing, generation settings, and export — with real prompt examples you can copy and use immediately.

Free to start · 1 credit on signupNo technical setup requiredWorks in any modern browser
Getting Started

How to Use Ernie Image — 4 Steps

Follow these steps in order. Each one takes under a minute.

  1. Create Your Account

    Sign up takes under a minute. Go to the Ernie Image homepage and choose how you want to register:

    Continue with Google

    One click — no form needed

    Sign Up with Email

    Enter your address and password

    Either way, every new account receives 1 free generation credit automatically on signup — enough to generate your first image immediately, at no cost.

    Note on free credits: The signup bonus applies to the first three accounts created from the same IP address. On shared networks — university, office, co-working space — the bonus may already be claimed. Purchase credits directly if that's the case.

    Once you're in, you'll see the generation interface. You're ready to start.

  2. Write Your Prompt

    In the Ernie Image AI image generator, the prompt is the only required input. Everything else has sensible defaults. Type your description in the prompt field — plain language works fine, and Ernie Image's built-in Prompt Expansion will automatically enrich your description before the model generates.

    You can be brief:

    a rainy street in Tokyo at night, cinematic lighting

    Or highly structured. For layout-sensitive content — posters, infographics, grid compositions — the more explicit your structure, the more precisely Ernie Image follows it.

    The Prompt Structure Formulas below show the exact building blocks for four common use cases. Fill in the blue-outlined slots with your own content and the fixed teal tokens handle the grammar and structure for you. For broader prompt principles, see the Prompt Writing Guide further down.

    Prompt Structure Formulas

    Each formula below shows the building blocks of a well-structured prompt. Teal = fixed structure word · Blue dashed = fill in your content · Grey = optional

    Poster or Grid Layout

    Educational charts, alphabet posters, infographic grids, product cards

    Composition
    [aspect ratio]composition.[background color]background.
    Style
    [visual style]style —[mood adjective]but[mood adjective].
    Title text
    [font style]title at[position]:"YOUR TITLE TEXT". Below it,[font style]subtitle:"subtitle text".
    Grid
    [N]rows ×[N]columns grid. Each cell contains:[element 1],[element 2],[label]below.
    Finish
    [color palette]colors throughout.[line quality]line weight. All textclearly legible.

    Photography & Cinematic Scene

    Mood shots, travel photography, atmospheric storytelling, editorial

    Shot type
    Photograph taken[camera position / angle].
    Foreground
    Foreground:[key object or framing element].
    Main subject
    Main view:[scene description].[sky / lighting condition]transitioning from[color A]to[color B].
    Lighting
    [light source and direction].[contrast quality]contrast.
    Mood & style
    [mood adjective],[emotional tone]mood.[film / aesthetic style]aesthetic.
    Text overlay (optional)
    White text overlaid[position]:"your text here"

    Mixed Media & Composite

    Illustration inside photo, social content with sticker elements, layered scenes

    Concept
    [creative concept name]composition blending[medium A]with[medium B].
    Container
    [object acting as frame]positioned[position / angle]on[surface].
    Inner element
    Inside / on screen:[style]illustration of[character description]. She / He[action that bridges inner and outer].
    Surrounding
    Around it:[decorative elements — emojis, stickers, text labels].
    Background
    Background:[real-world objects and their positions].[lighting quality]lighting.[aesthetic reference]aesthetic.

    Scientific & Multi-Panel Infographic

    Educational diagrams, concept explainers, data narratives, bilingual content

    Canvas
    [wide / square]-format[content type]infographic.[background color]backgroundwith [texture][render style].
    Title
    Top center:[font style]title"Main Title". Below it,[color]subtitle:"Subtitle text".
    Left panel
    Left panel —"Panel Name":[visual element + description]. Right of it:[text labels].
    Center panel
    Center panel —"Panel Name":[visual element]. Below it:[text box style]text box with[color]text:"Key statement."
    Right panel
    Right panel —"Panel Name":[sequence of visual elements with arrows]. Below:[explanatory text lines].
    Finish
    Balanced layout. Clear typographic hierarchy. No text overlap.[material / glow effects].[visual quality descriptor]clarity with visual elegance.
  3. Generate

    Once your prompt is ready, click Generate. Ernie Image submits your request and processes it in the background — most generations complete within 15–30 seconds depending on your settings and current queue load.

    Your results appear in two places:

    Right preview panel

    Visible immediately on the generation screen, alongside your settings.

    User Center — History

    All past generations are saved here automatically with their original prompts and settings.

    If you enabled batch generation (2–4 images), all results from the same prompt appear together so you can compare variations before deciding which to use.

    Credit usage: Credits are deducted when generation completes successfully. ERNIE Image costs 4 credits per image; Turbo costs 1 credit per image. If generation fails, no credits are charged. See Ernie Image pricing plans for credit pack options and per-image cost breakdowns.

    Generation Settings Reference

    SettingRangeDefaultWhat It Controls
    ModelERNIE Image / TurboERNIE ImageQuality vs. speed trade-off
    Aspect Ratio1:1, 16:9, 9:16, 4:3, 3:4, Custom1:1Output dimensions
    Custom Size64–2048 px per sideExact pixel dimensions
    Inference Steps1–10050Detail depth and generation time
    Guidance Scale1–204How strictly the model follows your prompt
    Num Images1–41Images generated per request
    Prompt ExpansionOn / OffOnAuto-enriches prompt via LLM before generating
    Safety CheckerOn / OffOnFilters NSFW content from output
    Output FormatPNGPNGFixed — all outputs are PNG
    When to use Turbo: Use ERNIE Image Turbo (1 credit, 8 steps) for rapid ideation — when you need to test 10 compositional ideas quickly. Switch to ERNIE Image (4 credits, 50 steps) for the final render where text accuracy and layout precision matter.
  4. Download and Access Your History

    When your images appear, click the Download button to save them to your device as PNG files. You can download individual images or all results from a batch at once.

    To revisit a previous generation, open the User Center. Every image you've generated is stored there with its original prompt and settings — you can re-download at any time or use a past prompt as the starting point for a new generation.

Prompt Writing Guide

How to Write Better Prompts for Ernie Image

Getting the most from Ernie Image comes down to how you structure your prompt. These principles apply across use cases — poster, photograph, or detailed infographic.

Start with the Visual Frame

Lead with the overall composition and format before describing the subject. Ernie Image interprets structural framing first.

a girl in a spacesuit
Square composition, flat vector illustration style. Center frame: a young girl in a white spacesuit with a transparent helmet, standing upright.

Describe Layout for Structured Content

For posters, infographics, and multi-panel images, specify the grid structure explicitly. Ernie Image handles layout-sensitive generation better than most open-weight models — but only if you tell it what the layout should look like. Include:

  • ·Number of rows and columns
  • ·Relative position of text and image elements
  • ·Font style and size hierarchy
  • ·Color palette and background color

Use Style Keywords Precisely

Ernie Image responds to specific visual style descriptors. Use precise terms rather than vague ones:

Instead ofUse
realisticphotorealistic, f/2.8 aperture, natural light
cartoonflat vector illustration, clean line weight, limited palette
darkcinematic, low-key lighting, deep shadow, chiaroscuro
colorfulbright saturated palette, bold primary colors
professionaleditorial layout, white background, minimal visual noise

Bilingual Text in Images

Ernie Image renders both English and Chinese text within the same generated image — one of its documented strengths. When your prompt includes in-image text:

  • ·Wrap the exact string in quotes: title: "ALPHABET OF CAREERS"
  • ·Specify font style if it matters: rounded bold lettering, clean sans-serif
  • ·Specify position: centered at top, overlaid in the lower third

Mixing languages works: heading: "职业字母表" subtitle: "A Guide to Different Jobs" will render both in the same image.

Prompt Expansion — When to Toggle It

The built-in Prompt Expansion LLM rewrites your input into a richer structured description before generation. Leave it on by default — turn it off only when precision matters more than richness.

Leave ON when…
  • Your prompt is short or rough
  • You want free visual interpretation
  • You're in ideation mode
Turn OFF when…
  • Prompt is already highly detailed
  • You need exact text placement
  • Iterating on a working prompt

Common Patterns by Use Case

Social media & marketing

Describe the crop ratio first (landscape 16:9 composition), then subject, then lighting and mood. Keep it under 80 words. Enable Prompt Expansion.

Posters with readable text

Describe background first, then title text in quotes with font style, then body layout. Use 1:1 or 4:3 ratio. Minimum 50 inference steps. Disable Prompt Expansion if you need exact wording.

Concept art & storyboards

Lead with mood and lighting (cinematic, overcast sky, rim lighting from left), then subject and action. Use Turbo for rapid variation, ERNIE Image for final renders.

Scientific & educational diagrams

Describe sections left to right or top to bottom. Label each visual element explicitly. Specify text placement and language. A guidance scale of 8–12 helps the model stay on brief.

FAQ

Ernie Image — Frequently Asked Questions

Do I need an account to use Ernie Image?

Sign up takes under a minute via Google or email. Every new account receives 1 free credit on signup — enough to generate your first image immediately at no cost.

What happens if I run out of credits?

Generation stops until you purchase more. Credits are available as one-time packs — Starter ($9.9 / 396 credits), Standard ($29.9 / 1,300 credits), and Pro ($49.9 / 2,626 credits). All credits are permanent and never expire.

Why did my image not look like my prompt described?

Most mismatches come from prompts that are too abstract. Add structure — describe the composition, layout, and style explicitly before describing the subject. For text-heavy output, turn Prompt Expansion off and increase Inference Steps to 70–100.

Can I use generated images in commercial projects?

Yes. The underlying ERNIE Image model is Apache 2.0 licensed. Generated outputs can be used in ads, products, client deliverables, and print without a separate commercial license.

Does Ernie Image offer an API?

No. Generation is available through the web interface only. For programmatic access to the underlying model, the open-source weights are available on Hugging Face under Apache 2.0.

Evaluating whether Ernie Image suits your needs? Read the Ernie Image review for a full quality assessment, pros and cons, and comparison with alternatives.

Start Generating

Ready to Generate?

Sign up free and open the Ernie Image generator in seconds. Every new account receives 1 free credit on signup — enough to generate your first image right away.