Wan 2.5 API Review: Current Integration Notes, Pricing, and Veo 3 Comparison

TL;DR

• Wan 2.5 is still the clean default when the job is short-form video volume, not campaign-grade cinematic work.
• On the current Evolink pricing surface, Wan 2.5 is $0.0708/sec, which is the same standard per-second band as the main Wan 2.6 tier on this site.
• The current repo positions Wan 2.5 around text-to-video and image-to-video, with short durations and predictable unit economics.
• If your product needs reference video, longer narrative control, or a more explicitly cinematic workflow, Wan 2.6 is the better fit.

What Wan 2.5 looks like in the current stack

• You care more about repeatable daily output than about squeezing every last bit of frame polish from a single render.
• You want a model that fits UGC pipelines, social content calendars, and productized video features.
• You need costs that are easy to explain to product and finance because the workload may scale to hundreds or thousands of clips.

The repo currently exposes Wan 2.5 in two practical modes:

• Text-to-video via wan2.5-text-to-video
• Image-to-video via wan2.5-image-to-video

Those flows run through the site's current unified video endpoint:

POST https://api.evolink.ai/v1/videos/generations

That is the integration surface developers should anchor to in this codebase, not older provider-specific route formats.

Current supported workflow

1. Text-to-video

The current Wan 2.5 text-to-video docs in this repo describe a simple request shape built around:

• model: "wan2.5-text-to-video"
• prompt
• aspect_ratio
• quality
• duration
• optional prompt_extend
• optional callback_url

2. Image-to-video

The current image-to-video flow uses:

• model: "wan2.5-image-to-video"
• prompt
• image_urls
• quality
• duration
• optional prompt_extend
• optional callback_url

This is the better fit when your application already has a first-frame visual, product shot, key art frame, or character image that the clip needs to respect.

3. Async task model

• You submit a generation request
• You receive a task id
• You poll task status or consume a callback
• Generated links are time-limited, so you should persist results promptly

That task-oriented flow makes Wan 2.5 easier to operationalize inside a backend job queue or a product workflow than a UI-only creator tool.

Pricing review

On the current site pricing surface, the main Wan 2.5 route is:

• Wan 2.5: $0.0708 / sec

That value comes directly from the repo's current model data and matches the price emphasized across the new Wan pricing and comparison content.

Practical cost examples

At the current repo price:

• 5 seconds is about $0.354
• 10 seconds is about $0.708

For the broader Wan picture, the most important comparison is not "is Wan 2.5 absolutely cheap in a vacuum?" but:

• It stays in the same standard per-second band as the main Wan 2.6 tier on this site
• It avoids the extra complexity that comes with Wan 2.6 reference-video planning
• It is positioned for the daily workhorse role, not the premium campaign role

Integration example

The safest copy-paste starting point for this codebase is the current unified endpoint:

curl --request POST \
  --url https://api.evolink.ai/v1/videos/generations \
  --header 'Authorization: Bearer YOUR_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "wan2.5-text-to-video",
    "prompt": "A sleek sports car accelerating through a neon-lit city at night",
    "aspect_ratio": "16:9",
    "quality": "720p",
    "duration": 5,
    "prompt_extend": true
  }'

And for image-to-video:

curl --request POST \
  --url https://api.evolink.ai/v1/videos/generations \
  --header 'Authorization: Bearer YOUR_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "wan2.5-image-to-video",
    "prompt": "Subtle camera push-in on the product, soft lighting, clean commercial look",
    "quality": "720p",
    "duration": 5,
    "image_urls": [
      "https://your-cdn.example.com/product-shot.png"
    ]
  }'

From an engineering perspective, the main production recommendations are straightforward:

• persist the returned task id immediately
• add retries/backoff around polling
• cache repeated generations where your product can tolerate it
• store finished media promptly because generated links are not forever

Wan 2.5 vs Veo 3: where Wan 2.5 still wins

The useful comparison is not "which model sounds more advanced on a landing page?" It is "which one fits the workload with less friction?"

Wan 2.5 tends to be the better fit when:

• you are shipping short-form social-style clips
• you need stable daily throughput
• your team cares about predictable unit economics
• you want one of the simplest routes into the Wan family on Evolink

Veo-class models tend to make more sense when:

• you are optimizing for premium hero output
• you can tolerate higher cost per asset
• your workload is closer to campaign creative than to daily content operations

That is why Wan 2.5 still matters even after newer cinematic tiers arrive. It is not obsolete. It simply owns a different part of the workload map.

Where Wan 2.5 feels strong in practice

In the current product positioning, Wan 2.5 is strongest for:

1. UGC and social production Short clips, fast iteration, and frequent publishing.
2. Image-led creative refreshes Teams with existing product art or key visuals can animate them into short motion assets without rebuilding the whole workflow around a more complex video tier.
3. Embedded SaaS video features If your app turns user prompts or uploaded images into short clips, the workhorse tier is often a better default than the premium tier.
4. Budget-sensitive experimentation Wan 2.5 gives teams a way to validate whether users even want generated video before moving part of the workload upstream to Wan 2.6.

Limits and caveats

A few cautions are worth being explicit about:

• In this repo, Wan 2.5 is documented around 5s and 10s jobs, not long-form generation.
• If your workflow depends on reference video / identity carryover from an existing clip, Wan 2.5 is the wrong tool; that is Wan 2.6 territory.
• You should avoid making stronger claims than the current docs support. The current site positioning clearly leans on audio-visual output, but if you are writing product or legal copy, use the live docs/dashboard as your final source of truth for exact request fields and current provider behavior.

FAQ

Is Wan 2.5 free?

No. In the current site content, Wan 2.5 is positioned as a paid model. Evolink may offer credits for new accounts in current site messaging, but pricing and credit availability should always be checked against the live product surface.

What is the current Wan 2.5 API endpoint on Evolink?

For this codebase, the current integration surface is:

POST /v1/videos/generations

What durations are currently documented?

Is Wan 2.5 better than Wan 2.6?

What should I read next?

• For current pricing: Wan API pricing guide
• For tier choice: Wan 2.5 vs Wan 2.6
• For the current product page and playground: Wan 2.5 model page

Conclusion

If you are building a product or workflow that needs cinematic multi-shot storytelling or reference video, you should move up to Wan 2.6. But if the real job is "generate a lot of short clips reliably without turning every generation into a premium production decision," Wan 2.5 remains a strong default.