Seedance 1.5 Pro is Live Now!Try Now!
WaveSpeed.ai
DashboardExploreInspiration
Home/Explore/Wan 2.6 Models/alibaba/wan-2.6/text-to-image
text-to-image

text-to-image

Alibaba WAN 2.6 Text-To-Image Model For AI Image Generation

alibaba/wan-2.6/text-to-image

Alibaba WAN 2.6 Text-to-Image turns text prompts into AI-generated images with the WAN 2.6 model for on-demand image creation. Ready-to-use REST inference API, best performance, no coldstarts, affordable pricing.

width
height
If set to true, the prompt optimizer will be enabled.

Idle

An extreme close-up documentary shot of a human face in brutal Arctic cold, eyelashes completely frozen and coated in thick ice crystals, frozen breath crystallizing in the air, skin slightly red from negative 50°C temperatures, hyper-realistic cinematic lighting, shallow depth of field, every frost particle sharply detailed, realistic cold blue color tones, shot on an ARRI Alexa 65 with a macro lens, natural film grain, Netflix-style documentary realism.

Your request will cost $0.03 per run.

For $1 you can run this model approximately 33 times.

One more thing::

ExamplesView all

An extreme close-up documentary shot of a human face in brutal Arctic cold, eyelashes completely frozen and coated in thick ice crystals, frozen breath crystallizing in the air, skin slightly red from negative 50°C temperatures, hyper-realistic cinematic lighting, shallow depth of field, every frost particle sharply detailed, realistic cold blue color tones, shot on an ARRI Alexa 65 with a macro lens, natural film grain, Netflix-style documentary realism.
a small girl with black twin-tail hair, sitting with her legs drawn together in front of her, smoking a cigarette, angel wings attached to her back, gently fluttering, flat solid gray background, no gradient, uniform monochrome, 3D pixel art style, voxel art, blocky geometry, anime-style character design, stylized proportions, minimal facial detail, low-resolution yet three-dimensional pixels, minimalistic composition, quiet and subdued mood, slightly surreal atmosphere, cinematic framing, soft but gloomy lighting --ar 58:77 --video 1
Jumping wolf motif that is one colour. The wolf is in similar style as Jankovics Marcell's Fehérlófia. As the wolf body looks like as flames. the wolf, standing in a snowy mountain landscape, minimalist ink sketch style, black and white only, sharp eyes, calm but tense posture, hand-drawn animation look, no fur details, abstract form, high contrast, rough texture --ar 1:1
dark fantasy 1980s DVD screengrab of a crusader raising his sword in a traditional early middle ages church ar 3:2 --ar 1:1
A mix collage with travis Scott, diamond, concert, neons, scratch paper, lyrics on paper, Ferrari, money, and girls with a futuristic vibe
A modern tea shop interior, warm afternoon light, minimalist wood design, cinematic photography, medium shot, shallow depth of field, 35mm look, clean lines, natural shadows, soft highlights, cozy seating, neatly arranged tea bar, high detail Negative prompt: blurry, low-res, watermark, text, logo, cluttered background, overexposed, underexposed, distortion, fisheye, noise

README

Alibaba Wan 2.6 Text-to-Image

Wan 2.6 Text-to-Image is Alibaba’s text-to-image model for generating PNG images from a single prompt. It’s designed for practical creative workflows—concept art, product visuals, portraits, and stylized imagery—where you want strong prompt adherence plus flexible output sizing.

Key capabilities

  • Single-request (synchronous) generation Generate images and get results back in one request (no separate polling flow).

  • Flexible custom dimensions Choose width*height freely as long as the total pixel area stays within the allowed range and the aspect ratio is within [1:4, 4:1].

  • Prompt rewriting via prompt_extend Optional prompt expansion can improve results for short/simple prompts (at the cost of a few extra seconds).

  • Negative prompting for tighter control Use negative_prompt to suppress unwanted artifacts (e.g., “blurry, low quality, extra fingers”).

  • Batch generation (n images per run) Generate 1–4 images per request; n directly affects cost (price scales with image count).

  • Reproducibility with seed Set seed for more consistent outputs across runs (results may still vary slightly).

  • Optional watermark control Toggle an “AI-generated” watermark in the lower-right corner.

Parameters and how to use

  • prompt: (required) The text instruction describing the image you want.
  • negative_prompt: What you want to avoid in the output image.
  • size: Output resolution as width*height.
  • n: Number of images to generate (1–4).
  • prompt_extend: Enable/disable prompt rewriting.
  • watermark: Enable/disable watermark.
  • seed: Random seed for more consistent results.

Prompt

Write prompts that are concrete and visual:

  • Start with subject + setting + style: “A modern tea shop interior, warm afternoon light, minimalist wood design, cinematic photography”
  • Add composition/camera when you care about framing: “medium shot, shallow depth of field, 35mm look”
  • If you need cleaner outputs, pair with a short negative prompt: “blurry, low-res, watermark, extra limbs, distorted hands”
  • Prompt language: Chinese and English are supported; prompts longer than the documented limit may be truncated.

Other parameters

  1. prompt (required) The positive prompt text that describes the content, style, and composition.

  2. size Format: width*height.

    • For wan2.6-t2i, total pixels must be between 768×768 and 1440×1440, and aspect ratio must be within [1:4, 4:1]. Common starting points:
    • 1:1 → 1280*1280 or 1024*1024
    • 16:9 → 1280*720
    • 9:16 → 720*1280

  3. negative_prompt Short list of undesired attributes. Max length is limited by the upstream API (extra text may be truncated).

  4. enable_prompt_expansion

    • true (default): improves results for short prompts; adds latency (often a few seconds).
    • false: faster; you control all detail manually.

  5. seed Integer in [0, 2147483647]. Use a fixed seed to reduce randomness; omit for fresh variations each run.

After you finish configuring the parameters, click Run, preview the result, and iterate if needed.

Pricing

Each run cost $0.03.

Notes

  • If you turn on enable_prompt_expansion, expect slightly higher latency; it’s best when your prompt is short or under-specified.
  • The image URLs returned by the upstream API may be time-limited—download and store outputs promptly if you need to keep them.

Related Models