Wan 2.6 Reference Video API

Use wan 2.6 reference video when you want your next video to clearly feel like "you" without re-shooting, re-editing, or restarting from scratch.

Reference Videos

Upload 1-3 reference videos to extract character appearance and voice.
Input billing capped at 5s total.

Prompt

Describe the video you want to generate...

Estimated Cost

6 Credits/s

(Input + Output duration × Quality multiplier)

Sample Result

No sample available

Prompt

0 (suggested: 2,000)

Reference Videos *

Upload 1-3 reference videos (2-30s each, max 100MB). Input billing capped at 5s total.

Click to upload or drag and drop

Supported formats: MP4, MOV
Maximum file size: 100MB; Maximum files: 3

Aspect Ratio

Quality

Duration

Click Generate to see preview

History

Max 20 items

0 running · 0 completed

Your generation history will appear here

wan 2.6 reference video: Turn any clip into scroll-stopping content

Upload a reference video, describe what you want, and get fresh versions that still look like your original—perfect for ads, UGC, and creator series.

What "reference video" means for creators

A practical definition

wan 2.6 reference video is a creator-friendly way to generate a new video while using an existing clip as your visual anchor. Instead of starting from a blank idea every time, you start from what already works: a face, outfit, product, scene, or brand vibe. With wan 2.6 reference video, the reference clip guides the look and feel, while your text request guides the change—so you can keep familiarity and still ship something new.

See example outputs

What you can make

Creators and marketers use wan 2.6 reference video to turn one "hero" clip into many purpose-built assets. Think: the same person introducing three product angles, the same scene remade for different seasons, or the same UGC hook rewritten for different buyer personas. Because wan 2.6 reference video is designed around a reference, it fits repeatable content: weekly series, daily drops, and always-on ad testing where speed matters.

Generate a variation

When it's the best choice

wan 2.6 reference video shines when the original clip already performs well and you want more outcomes without more production. It's ideal for A/B testing, scaling creatives across channels, or rescuing a great concept that needs a new vibe. wan 2.6 reference video is also useful when "identity drift" is a deal-breaker—like creator content, spokesperson ads, or a brand mascot that must stay recognizable.

Start with a reference

Why teams choose reference video

Reference-first creation cuts rework, protects brand consistency, and makes scaling content realistic for social and ads.

Stop losing the "same person" feel

Audiences notice continuity fast. wan 2.6 reference video keeps your creator/product recognizable while you change the hook, mood, or story angle—so new posts still feel like the same series.

Create more tests with less effort

Ads need volume: more hooks, CTAs, and angles. wan 2.6 reference video lets you generate variations from one strong reference clip, test faster, and iterate without constant re-shoots.

Keep brand style consistent across channels

One campaign often needs many formats. Use a single "style anchor" reference, then create platform-ready versions for Reels, Shorts, TikTok, and X while keeping the same brand look.

How to use it (simple workflow)

Follow this quick flow to get results you can actually post using wan 2.6 reference video.

Add a reference clip

Upload or paste a reference video that clearly shows the person, product, and vibe you want to preserve.

Describe the change

Write one clear request: new hook, new setting mood, new pacing, or a new angle—keep it human and direct.

Generate, pick, and ship

Create multiple outputs, choose the best, then post or export for ads, scheduling tools, and client approvals.

Build your first workflow on evolink.ai

Features that map to real outcomes

Each feature is written from a creator's point of view: less rework, more output, and better consistency.

Consistency

Reference-first consistency

wan 2.6 reference video helps you keep the "this is the same creator/product" feeling so your audience recognizes you instantly. Use it for recurring series, spokesperson ads, and brand mascots where continuity matters more than flashy experiments.

Performance

Fast variation for UGC ads

Use wan 2.6 reference video to spin a single winning UGC clip into multiple hooks and CTAs. This supports practical testing: different offers, different pain points, and different audiences—without booking another shoot.

Brand

Campaign-safe brand look

wan 2.6 reference video is useful when you want every post in a campaign to "belong together." Start with one approved reference, then generate versions that match your feed, your landing page tone, and your paid social creative.

Ease of use

Creator-friendly prompting

wan 2.6 reference video works best with plain language. Instead of tweaking complicated options, focus on what viewers will notice: the opening line, the emotion, the setting, and the product moment.

Social

Social format flexibility

wan 2.6 reference video fits the reality of social distribution: you often need different cuts for different platforms. Generate options for vertical-first posting, and keep the message tight for quick attention windows.

Workflow

Iteration loop that scales

wan 2.6 reference video supports a repeatable system: reference → variations → publish → learn → improve. This is how creators and small teams stay consistent while increasing output without sacrificing quality.

How it compares (API pricing + practical fit)

Pricing is referenced from public fal/Replicate pages; specs may vary by plan and provider.

Model	Duration	Resolution	Price	Strength
wan 2.6 (reference video)	N/A	HD (provider-dependent)	Varies by provider	Reference-first workflow for consistent variations from an existing clip (best for UGC/ad iteration).
Kling 2.6 (Pro, fal.ai)	N/A	HD (not explicitly listed on fal model page)	$0.07/sec (audio off) or $0.14/sec (audio on)	Cinematic image-to-video with optional native audio; good when you want higher production feel and integrated speech.
Google Veo 3.1 Fast (Replicate)	N/A	720p–1080p (as listed for Veo on Replicate's model comparison)	$3.20 per 8s video (≈ $0.40/sec) for Veo 3 Fast on Replicate's comparison table	High-fidelity short-form generations with audio-oriented workflows; strong for premium concepting and fast iterations.

FAQ about reference video

Everything you need to know about the product and billing.

wan 2.6 reference video is a reference-first video generation workflow from Alibaba's Wan2.6 model family. You upload a short reference clip, then describe the change in plain English (new hook, mood, setting, or storyline) to generate a fresh version that still feels like the same person/product. In many API setups, you can use 1–3 reference videos and call them out in the prompt to keep identity and style consistent across multiple variations. It's mainly used to scale UGC ads and creator series faster, without losing the recognizable "same clip / same character" feel.

Use wan 2.6 reference video when you already have a clip that performs well or looks exactly on-brand. Text-to-video can be great for experimenting, but it can also feel unpredictable when you need a specific face, product, or setting to stay recognizable. If your content is series-based (daily tips, weekly reviews, recurring characters), wan 2.6 reference video is usually the more practical option because it starts from your proven visual identity. It's also ideal when you need multiple ad variations quickly without planning another shoot.

To get strong results with wan 2.6 reference video, pick a reference clip that clearly shows what must remain stable: the subject, the outfit, the product, and the general lighting. Avoid references that are extremely blurry or constantly changing scenes if your goal is identity consistency. Then write one focused request at a time—like "make it feel more premium," "make the hook more urgent," or "make it look like an outdoor morning scene." With wan 2.6 reference video, clarity beats complexity, and smaller changes often look more believable.

Yes—wan 2.6 reference video fits UGC ad testing because it supports volume without losing familiarity. A typical workflow is to keep the same reference clip (same creator and product), then generate multiple versions that change only what you want to test: the first sentence, the offer, the problem statement, or the ending CTA. This keeps your test cleaner because your visuals are not changing randomly. Many teams treat wan 2.6 reference video like a "creative multiplier" for paid social, especially when ad fatigue hits fast.

wan 2.6 reference video is best understood as "guided variation," not a perfect clone. The reference clip acts like an anchor so the new output can stay recognizable, but changes you request may still introduce differences in details, motion, or overall feel. If you need the output to match your original frame-by-frame, traditional editing is still the right tool. For most creators, the sweet spot is using wan 2.6 reference video for new angles and new hooks that still feel like the same series—close enough for trust, different enough for freshness.

wan 2.6 reference video can be very effective for campaign consistency if you treat the reference clip as your "style source of truth." Start with one approved reference that reflects your brand: product framing, creator tone, lighting, and overall vibe. Then generate the versions you need for each channel, keeping the same anchor while adjusting the message for that audience. Used this way, wan 2.6 reference video helps avoid the common problem where social posts feel disconnected from ads, and ads feel disconnected from landing pages.

For wan 2.6 reference video, prompts work best when they sound like instructions you would give an editor or a creator friend. Focus on viewer-visible outcomes: the opening hook, the emotion, the setting vibe, the product moment, and the pacing. If you want multiple versions, write prompts that define a single change per version, such as "more playful tone," "more premium and clean," or "more urgent, problem-first." wan 2.6 reference video is not about stacking fancy keywords; it's about making a clear creative decision and asking for it plainly.

evolink.ai is positioned as the place to run a practical wan 2.6 reference video workflow: upload a reference, generate multiple variations, then pick and ship the ones that match your goal. For creators, that might mean turning one strong clip into a week of posts. For marketers, it might mean producing a set of ad angles for testing while keeping the same recognizable spokesperson. The key benefit is reducing the "blank page" feeling—wan 2.6 reference video gives you a repeatable starting point so content creation feels like a system, not a struggle.

API Reference

Select endpoint

Authentication

All APIs require Bearer Token authentication.

Header

Authorization: 
Bearer YOUR_API_KEY

Get API Key

POST

/v1/videos/generations

Create Video

WAN 2.6 Reference Video (wan2.6-reference-video) model supports reference video-to-video generation, extracting character appearance and voice from uploaded reference videos.

Asynchronous processing mode, use the returned task ID to .

Generated video links are valid for 24 hours, please save them promptly.

Request Parameters

modelstringRequiredDefault: wan2.6-reference-video

Video generation model name.

Examplewan2.6-reference-video

promptstringRequired

Text description of the video to generate.

Notes

Maximum 1500 characters

ExampleA person dancing

video_urlsarrayRequired

Array of reference video file URLs. Used to extract character appearance and voice.

Notes

Maximum 3 videos
Format: mp4, mov
Duration: 2-30 seconds each
File size: max 100MB each
Input billing capped at 5s total

Example["https://example.com/reference.mp4"]

aspect_ratiostringOptionalDefault: 16:9

Video aspect ratio.

Value	Description
16:9	Landscape video (default)
9:16	Portrait video
1:1	Square video
4:3	Standard video
3:4	Portrait standard

Example16:9

qualitystringOptionalDefault: 720p

Video quality. Higher quality costs more.

Value	Description
720p	Standard quality (1.0x price)
1080p	High quality (1.67x price)

Example720p

durationintegerOptional

Duration of the generated video in seconds.

Value	Description
5	5 seconds
10	10 seconds

Notes

Price is calculated as: base_price × (input_duration + output_duration) × quality_multiplier
Input duration is capped at 5 seconds for billing

Example5

model_params.shot_typestringOptionalDefault: single

Shot type for video generation.

Value	Description
single	Single continuous shot
multi	Multiple camera angles/shots

Examplesingle

callback_urlstringOptional

HTTPS callback address after task completion.

Notes

Triggered on completion, failure, or cancellation
HTTPS only, no internal IPs
Max length: 2048 chars
Timeout: 10s, Max 3 retries

Examplehttps://your-domain.com/webhooks/video-task-completed

Request Example

{
  "model": "wan2.6-reference-video",
  "prompt": "A person dancing",
  "video_urls": [
    "https://example.com/reference.mp4"
  ],
  "aspect_ratio": "16:9",
  "quality": "720p",
  "duration": 5,
  "model_params": {
    "shot_type": "single"
  }
}

Response Example

{
  "created": 1757169743,
  "id": "task-unified-1757169743-abc123",
  "model": "wan2.6-reference-video",
  "object": "video.generation.task",
  "progress": 0,
  "status": "pending",
  "task_info": {
    "can_cancel": true,
    "estimated_time": 120
  },
  "type": "video",
  "usage": {
    "billing_rule": "per_second",
    "credits_reserved": 10,
    "user_group": "default"
  }
}