Wan 2.5 Image

Image GenerationText / Image InputUp to 4 Outputs

Wan 2.5 Image is APIXO’s still-image route for Alibaba’s Wan 2.5 generation and editing workflows. Create images from text or transform one to three references with prompt and negative-prompt guidance, mode-specific aspect ratios, and one to four requested variants per asynchronous task.

Workflows

Text / Image to Image

APIXO Price

$0.03 / Image

References

Up to 3 Images

Output Count

1–4 Images

Prompt

Up to 2,000 Characters

Capabilities

Explore concepts, edits, and reference combinations

Prompt-Led Generation

Describe subjects, settings, composition, lighting, materials, and visual style in Chinese or English to produce new still-image concepts without a source image.

Instruction-Based Editing

Submit existing imagery with a required instruction to transform its content or appearance. The workflow does not expose masks, layers, or bounding-box controls through APIXO.

Multi-Image Fusion

Combine visual information from up to three references for composition, subject, or style development, while allowing for variation in identity and reference relationships.

Negative Prompting

Add a negative prompt of up to 500 characters to discourage unwanted objects, visual defects, or stylistic traits during generation.

Variant Exploration

Request one to four alternatives in a task to compare interpretations, compositions, and treatments without assuming that every requested image will complete successfully.

Format-Aware Framing

Choose common square, portrait, or landscape ratios in both modes, with additional 2:3, 3:2, and 21:9 framing available for image-to-image.

Generation modes

Choose a blank canvas or reference-guided transformation

Text-to-Image

Start with a required, non-empty prompt and no source images. APIXO supports 1:1, 3:4, 4:3, 9:16, and 16:9 framing, plus one to four requested outputs and an optional negative prompt.

Prompt → Images

Image-to-Image

Provide one to three public images and a required instruction for editing, transformation, or fusion. This mode supports the five standard ratios plus 2:3, 3:2, and 21:9, but it does not guarantee exact identity, geometry, layout, text, or logo preservation.

References → Variants

Route details

Wan 2.5 Image specifications

Current inputs and output controls documented for APIXO’s still-image route.

2 Modes

Workflows

2,000 Characters

Prompt Limit

500 Characters

Negative Prompt

1–3 URLs

References

1–4

Requested Images

Async / Callback

Delivery

Production uses

Develop visual directions across practical workflows

Campaign ideation

Explore Creative Variants

Generate several interpretations of a campaign brief for mood boards, pitch decks, and early art direction. Compare outputs for composition, style, and brand fit before moving selected concepts into production.

Product imagery

Develop Ecommerce Scenes

Use product references to explore alternate settings, lighting, and promotional compositions. Review proportions, materials, labels, logos, and small construction details against approved product assets before publishing.

Visual development

Combine Reference Directions

Bring subject, composition, color, or stylistic cues from several images into one exploratory result. Use clear prompt language to explain each reference’s intended role and expect some compositional drift.

Design exploration

Test Posters and Key Art

Iterate on poster layouts, illustration treatments, social creative, and key-art concepts across portrait, landscape, and ultrawide formats. Add final typography and exact brand marks in a conventional design tool.

Notes and FAQ

Input rules, billing, and review guidance

Before you generate

wan-2-5-image is a still-image route only; it does not include Wan 2.5 video generation, audio, duration, or motion controls.

APIXO exposes text-to-image and image-to-image under one product route but does not publicly confirm its internal mapping to Alibaba’s wan2.5-t2i-preview and wan2.5-i2i-preview IDs.

Image-to-image requires one to three public JPEG, JPG, non-alpha PNG, BMP, or WebP URLs with width and height between 384 and 5,000 pixels. Keep every URL directly accessible while the task runs.

APIXO recommends the first status poll after approximately 10–15 seconds. Continue with a moderate polling interval or use callback delivery; no fixed generation-time SLA is published.

Successful image URLs are stored in resultJson.resultUrls. A task can succeed with fewer URLs than requested when only part of the batch completes.

Review identity, anatomy, hands, object relationships, product proportions, embedded text, logos, reference fidelity, and unwanted additions. Use only source material you have permission to process.

Frequently asked questions

APIXO’s detailed documentation lists $0.03 per successfully generated image for both modes. One completed image costs $0.03, while four completed images cost $0.12. Partially completed tasks are billed according to the successful outputs returned.

No. Four is the API default when num_images is omitted, not a completion guarantee. You may request one, two, three, or four images, and a partially successful task can return fewer URLs than requested.

APIXO documents the same per-successful-image rate for text-to-image and image-to-image. No separate public fee is listed for reference count, aspect ratio, or callback delivery.

APIXO exposes aspect-ratio selection but does not document custom pixel dimensions, 1K or 2K presets, transparent output, or a selectable file format on this route. Alibaba identifies PNG as the upstream output format, but APIXO returns result URLs rather than a format control.

No. References can guide editing and fusion, but exact faces, product geometry, composition, text, logos, and style relationships are not guaranteed. Deterministic seed control, masks, inpainting, outpainting, and formal subject- or style-reference modes are not exposed by APIXO.