Wan 3.0 API: Complete Pricing and Feature Guide (2026)

paymentsComplete Pricing Table

Three resolution tiers, all covering the same 2–30 second duration range:

Billing is per second of generated video, pay-as-you-go — the price is the same whether native sound is on or off. There is no subscription and no minimum deposit; a $10 top-up (the minimum, via PayPal) covers forty seconds of 480p output. Every clip renders as a 30fps MP4 in five aspect ratios: 16:9, 9:16, 1:1, 4:3 and 3:4.

layersCreation Modes: One Model, Three Inputs

Text-to-video. A pure prompt produces the clip, with native audio — ambience, sound effects, optional dialogue — generated alongside the picture rather than bolted on afterward.

Image-to-video with first/last-frame control. Provide a first-frame image and Wan 3.0 animates from it. Add an optional last-frame image and the model interpolates the motion between the two — the cleanest way to produce loopable shots. The last-frame trick is also how you chain 30-second takes into longer sequences with seamless continuity.

Reference-to-video. The mode that separates Wan 3.0 from most of the field: attach up to 10 reference images, 5 reference videos and 5 reference audio tracks, and the model grounds the generation in your material — preserving a character's look across shots or a product's exact appearance. No other flagship on Viddra accepts this input mix.

scheduleThe 30-Second Advantage

Most frontier models cap a single generation at 4–15 seconds. Wan 3.0 runs 2 to 30 seconds in one pass — among the longest single-shot durations available in 2026 — which changes the economics of anything longer than a social loop: one generation instead of three, no stitching artifacts, no continuity drift between segments.

compareWan 3.0 vs Veo 3.1: The Price Math

Veo 3 starts at $0.35/sec and Veo 3.1 Fast runs a flat $0.24/sec — nominally cheaper per second than Wan 3.0's 720p tier ($0.50) and close to its 480p tier ($0.25). But Veo clips cap at roughly 8 seconds, so a 30-second sequence on Veo 3 means four generations and a join: about $10.50 of footage with visible seams to manage. Wan 3.0 renders the same 30 seconds in one pass at $7.50 (480p) or $15.00 (720p), and adds reference-media input that Veo does not offer.

The practical split: short voiced b-roll on a budget goes to Veo 3.1 Fast; long-form sequences, brand-consistent characters and first/last-frame control go to Wan 3.0. Our Veo 3.1 access guide covers the Google side in detail.

terminalcurl Quickstart

Text-to-video in one call — POST to /video/generations with model wan3.0-video:

curl -X POST https://api.viddra.com/v1/video/generations \
  -H "Authorization: Bearer $VIDDRA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "wan3.0-video",
    "prompt": "Camera slowly pushes forward over a misty mountain lake at golden sunrise, birds crossing the frame, cinematic documentary style",
    "duration": 10,
    "aspect_ratio": "16:9"
  }'

# 202 Accepted
# { "task_id": "task_...", "status": "queued" }

Poll the task endpoint until the status is succeeded, then download the MP4 from video_url — the same async pattern every Viddra video model uses. Image-to-video and reference-media request fields are documented in the API documentation.

helpFrequently Asked Questions