Imagen4

Imagen4

Playground

Imagen 4 is an AI image generator that produces detailed, high-quality images up to 2K across diverse styles and subjects. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

Features

Imagen 4 (Text-to-Image) — wavespeed-ai/imagen4

Imagen 4 is a high-quality text-to-image model for generating photorealistic and stylized images from prompts, with optional negative prompts for extra control. It’s well-suited for concept art, marketing creatives, portraits, and detailed scenes where you want clean composition and strong realism.

Key capabilities

• Text-to-image generation with strong realism and detail
• Optional negative_prompt to reduce unwanted artifacts or styles
• Reliable composition across common aspect ratios (e.g., 1:1, 16:9, 9:16)
• Supports multi-image output in a single run via num_images

Pricing

$0.04 per image.

Total cost = num_images × $0.04 Example: num_images = 4 → $0.16

How to use

1. Write a prompt describing subject, setting, lighting, and style.
2. Optionally add a negative_prompt listing what you want to avoid.
3. Select aspect_ratio for your target layout.
4. Choose num_images for how many variations you want.
5. Set seed for reproducible results (optional), then generate.

Parameters

• prompt (required): The text description of what to generate
• negative_prompt (optional): What to avoid (artifacts, styles, objects, text, etc.)
• aspect_ratio: Output aspect ratio (e.g., 1:1, 16:9, 9:16)
• num_images: Number of images to generate per run
• seed: Fixed value for reproducibility; leave empty/random for variation

Prompting guide

A stable structure:

• Subject: who/what is in frame
• Scene: where + time + atmosphere
• Details: clothing, materials, environment cues
• Lighting: softbox, sunset, harsh noon, rim light
• Camera: close-up, wide shot, depth of field
• Style: photorealistic, cinematic, illustration, etc.

Example pattern: A [shot type] of [subject] in [scene]. [Lighting + mood]. [Camera cues]. [Style cues].

Example prompts

• A raw, unflinching photograph of a weathered soldier in a desert trench, dust blowing across his helmet and gear, harsh sunlight, shallow depth of field, cinematic realism.
• Studio product photo of a minimalist watch on a matte surface, softbox lighting, crisp shadow, premium advertising look.
• Cozy café interior at night, warm tungsten lighting, rain on windows, film still composition, subtle grain.

Negative prompt examples

• blurry, low resolution, deformed hands, extra fingers, bad anatomy
• watermark, logo, text, caption, jpeg artifacts
• oversaturated, cartoonish, plastic skin, uncanny face

Best practices

• Use negative_prompt sparingly; focus your main prompt on what you want.
• Keep the first sentence concrete, then add camera/lighting constraints.
• Fix seed when iterating on prompt wording for controlled comparisons.

# Submit the task
curl --location --request POST "https://api.wavespeed.ai/api/v3/wavespeed-ai/imagen4" \
--header "Content-Type: application/json" \
--header "Authorization: Bearer ${WAVESPEED_API_KEY}" \
--data-raw '{
    "aspect_ratio": "1:1",
    "num_images": 1
}'

# Get the result
curl --location --request GET "https://api.wavespeed.ai/api/v3/predictions/${requestId}/result" \
--header "Authorization: Bearer ${WAVESPEED_API_KEY}"