
COMMUNITY PAGE
Wan 3: 30 seconds of video with sound, in one run
Run Wan 3 workflows with no installs
alibaba
text to video
video with audio
wan 3.0 prime
Generate a high-fidelity clip with sound from a written prompt using Wan 3.0 Prime, the quality tier of Alibaba's latest video model. No image needed, just describe the shot.
Wan 3.0 Prime · Text to Video With Audio
Generate a high-fidelity clip with sound from a written prompt using Wan 3.0 Prime, the quality tier of Alibaba's latest video model. No image needed, just describe the shot.
multi-reference
reference to video
video to video
wan 3.0 prime
Place a character into any video using Wan 3.0 Prime by Alibaba. Upload a face, a reference clip, and describe the scene. Prime fidelity, audio included.
Wan 3.0 Prime · Reference to Video
Place a character into any video using Wan 3.0 Prime by Alibaba. Upload a face, a reference clip, and describe the scene. Prime fidelity, audio included.
alibaba
image to video
video with audio
wan 3.0 prime
Animate a still into a high-fidelity clip with sound using Wan 3.0 Prime, the quality tier of Alibaba's latest video model. Upload an image, describe the shot.
Wan 3.0 Prime · Image to Video With Audio
Animate a still into a high-fidelity clip with sound using Wan 3.0 Prime, the quality tier of Alibaba's latest video model. Upload an image, describe the shot.
alibaba
video style transfer
video to video
wan 3.0
Re-render any video in a new visual style using Wan 3.0 by Alibaba. Upload a clip, describe the target look, and hit run. Motion preserved, audio included.
Wan 3.0 · Reference to Video
Re-render any video in a new visual style using Wan 3.0 by Alibaba. Upload a clip, describe the target look, and hit run. Motion preserved, audio included.
What you get
Wan 3 generates up to 30 seconds of video with sound, in a single run.
You get it as three nodes: Text-to-Video, Image-to-Video, and Reference-to-Video.
Text-to-Video takes a prompt. Image-to-Video takes a start frame, and an optional last frame, so you can pin both ends of a shot and let the model solve the middle. Reference-to-Video takes 10 images, 5 videos and 5 audio clips in one generation, so a character, a style, a movement and a voice can all hold inside the same shot. References run up to 15 seconds total across all of them.
Every feature
Standard and Prime
Every node comes in two tiers. Prime is the same model at the same quality, and it comes back about seven times faster.
Which one to run
Run Prime while you are still deciding, when the next thing you do depends on seeing this one. Run Standard when you are batching overnight, sweeping seeds, or otherwise not watching.
Find the shot on Prime at 720p. Commit the final at 1080p on Standard. That way you only pay the premium while you are actually waiting on the result.
What it costs
You pay per second of finished video, not per second of render.
| Per second | 480p | 720p | 1080p |
|---|---|---|---|
| Standard | $0.05 | $0.10 | $0.20 |
| Prime | $0.068 | $0.14 | $0.28 |
Per shot:
| Per shot | 480p Std | 720p Std | 1080p Std | 480p Prime | 720p Prime | 1080p Prime |
|---|---|---|---|---|---|---|
| 5 seconds | $0.25 | $0.50 | $1.00 | $0.34 | $0.70 | $1.40 |
| 10 seconds | $0.50 | $1.00 | $2.00 | $0.68 | $1.40 | $2.80 |
| 15 seconds | $0.75 | $1.50 | $3.00 | $1.02 | $2.10 | $4.20 |
| $4.20 | $8.40 |
How it compares
| Spec | Wan 3 | Wan 2.7 | Wan 2.2 |
|---|---|---|---|
| Max duration | 30s | 15s | 5s |
| Max duration, reference-to-video | 30s | 10s | n/a |
| Max resolution | 1080p | 1080p | 720p |
| Native audio | Yes, default on | Yes | No |
| Reference assets | 20 | 5 combined | None |
| LoRAs | No | Yes | Yes |
Wan 3 reference assets: 10 images, 5 videos, 5 audio. Wan 2.2 and 2.7 figures are the published model specs. Wan 2.2 is open weights, so community builds routinely push past them.
What it does not do yet
Wan 3 does not support LoRAs, and Alibaba has not given a date.
Twenty references approximate a trained likeness rather than guarantee it. If your pipeline depends on one, Wan 2.2 with Wan Animate is still the better tool for that step. We will update this page the day LoRA support lands.
The timings above are 720p at 15 seconds. Other resolutions, 30 second shots, and behavior under load will differ. We will publish those once enough runs sit behind them to mean something.
The six nodes, by name
Alibaba Wan 3.0 on Floyo. Up to 30 seconds of video with native audio in one run, at up to 1080p, across three nodes in a standard and a high-speed tier. No LoRA support yet.