Alibaba model: frame-to-frame transitions, custom audio track overlay (speech-to-video), reference video generation, and editing of finished clips.
In the studio, click the model name at the top, then find on the left Wan — versions on the right:
Wan 2.7 parameters:
Wan 2.7's standout feature — video synced to YOUR audio track:
First frame
Last frameWan performs most consistently with a clear scene structure.
Hit "Generate" — the task enters the background queue (usually 1–3 minutes). The finished video will appear in your feed and in "My Videos"; tokens for failed generations are refunded automatically.
Real results from Airium Studio — hit play:
How to generate video in Wan 2.7 online?
Open Airium Studio, select the Wan 2.7 model in the catalog, set parameters, write your prompt, and click "Generate". Registration takes a minute — new users receive free tokens.
How much does video generation cost in Wan 2.7?
In Airium Studio, generating with Wan 2.7 costs approximately 22 G-tokens / 5 sec. You only pay for successful generations — tokens for failed ones are returned automatically.
What is the maximum video duration in Wan 2.7?
Available durations: 4–15 s (2.7) · 5/10 (2.5). Parameters are set directly in the studio before generation.
Can I use Wan 2.7 without API keys or VPN?
Yes — Airium Studio works in the browser with no API keys, foreign bank cards, or VPN: simply select Wan 2.7 in the catalog and generate online.
| Parameter | Values | Tip |
|---|---|---|
| Duration | 4–15 s (2.7) · 5/10 (2.5) | whole seconds |
| Frames | first + last (2.7) | 2.5 supports start frame only |
| Audio | custom track: t2v background · i2v lip-sync | record in "Voiceover" |
| Reference video | motion dynamics transfer | r2v mode |
| Edit | editing a finished video | describe the changes in a prompt |
| Control | negative prompt + seed | reproducibility |
Photo + your audio track = a character delivers the text with lip-sync.
First and last frame + description of the connection between them.
negative prompt: «blur, artifacts, extra fingers, text».
One seed — change only the prompt wording and compare results.