WAN 3 Reference
WAN 3 Reference costs $0.24/sec on FairStack · a image to video model for Character or product consistency across shots, Style-locked sequences, Turning a set of stills into one coherent clip. No subscription required. Pay per generation with full REST API access. FairStack applies a transparent 20% margin on infrastructure cost so you always see the real price.
What is WAN 3 Reference?
WAN 3.0 Reference-to-Video generates a clip guided by reference images, keeping subjects and style consistent while following your prompt. Reference the images positionally in the prompt — for example, 'the subject in Image 1 walks past the building in Image 2'. Up to nine reference images per generation, at 480p, 720p or 1080p, with audio generated in the same pass. Billing is per second of output, at the model's cost plus a flat 20% platform fee, shown before you spend. Upstream WAN 3 accepts up to 30 seconds in one generation; on FairStack it is capped at 16 seconds, the platform-wide ceiling.
Key Features
What are WAN 3 Reference's strengths?
What are WAN 3 Reference's limitations?
What is WAN 3 Reference best for?
How much does WAN 3 Reference cost?
How does WAN 3 Reference perform across capabilities?
Estimated from the WAN 3 family — NOT independently benchmarked. Re-score after a Visual Observatory sweep. PRD-1028.
How do I use the WAN 3 Reference API?
curl -X POST https://api.fairstack.ai/v1/generations/video \
-H "Authorization: Bearer $FAIRSTACK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "wan-3-r2v",
"prompt": "Your prompt here"
}'import requests
response = requests.post(
"https://api.fairstack.ai/v1/generations/video",
headers={
"Authorization": f"Bearer {FAIRSTACK_API_KEY}",
"Content-Type": "application/json",
},
json={
"model": "wan-3-r2v",
"prompt": "Your prompt here",
},
)
result = response.json()
print(result["url"])const response = await fetch(
"https://api.fairstack.ai/v1/generations/video",
{
method: "POST",
headers: {
Authorization: `Bearer ${process.env.FAIRSTACK_API_KEY}`,
"Content-Type": "application/json",
},
body: JSON.stringify({
model: "wan-3-r2v",
prompt: "Your prompt here",
}),
}
);
const result = await response.json();
console.log(result.url);What parameters does WAN 3 Reference support?
Frequently Asked Questions
How much does WAN 3 Reference cost?
WAN 3 Reference costs $0.24/sec on FairStack as of 2026-09-24. This price includes FairStack's transparent 20% margin on infrastructure cost. No subscription or monthly fee — you pay per generation only. Minimum deposit is $1.
What is WAN 3 Reference and what is it best for?
WAN 3.0 Reference-to-Video generates a clip guided by reference images, keeping subjects and style consistent while following your prompt. Reference the images positionally in the prompt — for example, 'the subject in Image 1 walks past the building in Image 2'. Up to nine reference images per generation, at 480p, 720p or 1080p, with audio generated in the same pass. Billing is per second of output, at the model's cost plus a flat 20% platform fee, shown before you spend. Upstream WAN 3 accepts up to 30 seconds in one generation; on FairStack it is capped at 16 seconds, the platform-wide ceiling. WAN 3 Reference is best for Character or product consistency across shots, Style-locked sequences, Turning a set of stills into one coherent clip. Available via FairStack's REST API with curl, Python, and Node.js SDKs.
Does WAN 3 Reference have an API?
Yes. WAN 3 Reference is available via FairStack's REST API at api.fairstack.ai. Send a POST request to /v1/generations/video with your API key and prompt. Works with curl, Python requests, Node.js fetch, and any HTTP client. No SDK installation required.
How does WAN 3 Reference compare to other video models?
WAN 3 Reference excels at Character or product consistency across shots, Style-locked sequences, Turning a set of stills into one coherent clip. It is a image to video model priced at $0.24/sec on FairStack. Key strengths: Subject and style consistency across a shot, Native synced audio. Compare all video models at fairstack.ai/models.