Wan 3.0 text-to-video: write the shot, then the settings
How to use Wan 3.0 text-to-video on wan-ai.app: prompt shape, 2–30 second lengths, 480P to 1080P, native audio, and credit cost.
Posted by
Related reading
Wan 3.0 image-to-video: first frame, optional last frame
How Wan 3.0 image-to-video works here: a required first frame, an optional last frame, file limits, and a prompt that describes motion.
Wan 3.0 reference-to-video, and where this site does it
Reference photos on wan-ai.app live in the templates, not on the first-frame tab. What that changes about motion, credits, and consent.
How to make a 30-second Wan 3.0 video without wasting the credits
Wan 3.0 can run 30 seconds in one submit. Credit math at 480P, 720P, and 1080P, and a shot list that survives the length.
Text-to-video on the Wan 3.0 generator is the mode with no upload. You describe the shot, pick a length from 2 to 30 seconds, and the clip comes back with optional native audio. This is the path to learn before you spend credits on a 30-second 1080P rerun.
Open the right tab
On /wan-3-0, stay on Text to Video. Image to Video is a different contract: it pins a first frame, and an optional last frame. If you only have words, text mode is the one that will run.
The form starts at 720P, 16:9, 5 seconds, audio on. Those defaults are a reasonable first paid test. A 5-second 720P Standard clip is 3 credits. Change them before you submit if you want a vertical draft or a silent picture.
Write the shot, not a topic
A topic (“a sad city”) gives the model room to invent a poster, a caption, and a second subject you did not ask for. A shot names one subject, the place, the light, the camera, and the sound. Four beats is enough for an 8-second clip:
- Who is in frame, and that there is only one of them if that matters.
- Where the camera sits, and whether it moves.
- What changes between the start and the end. If nothing should change, say hold.
- What the microphone hears, including “no speech” when you want quiet.
Put the refusals in the same prompt. This generator has no negative-prompt box. “No titles, no logos, no readable signs” belongs in the text, not in a second field that does not exist.
The prompt library has original examples in this shape, including a night platform and a kitchen window. Use this prompt on a card opens the generator with that text and the suggested settings filled in. Edit the length before you spend the credits.
Settings that change the bill
Resolution is 480P, 720P, or 1080P. Aspect is 16:9, 4:3, 1:1, 3:4, 9:16, or adaptive. Adaptive is aimed at image jobs, where the still can suggest the frame. For text, pick the frame yourself. A vertical ad is 9:16. A film beat is 16:9. A packshot often wants 1:1.
Duration is any whole second from 2 to 30. The slider and the 5, 10, 15, and 30 buttons set the same value. There is no “smart duration” on this page, so a short prompt does not silently become a long clip, and a long prompt does not get cut to 5 seconds unless you left the duration there.
Standard credits round up: about 1 credit per 4 seconds at 480P, per 2 seconds at 720P, and per 1 second at 1080P, with a minimum of 1 credit. A 5-second 720P Standard clip is 3 credits. An 8-second 720P test is 4. A 30-second 1080P Standard clip is 30. Prime is the faster model and costs more for the same settings. The button shows the number for the tier you have selected. If the job fails to submit or the model fails, those credits are refunded.
Audio is part of the prompt
Native audio is on by default. If the prompt never mentions sound, you still get a soundtrack, and it may include a voice you did not plan. Say what the mic hears. Rain and a brake squeal. A pour and a drip. Or “no speech, no music” when the picture should carry the clip.
Turn audio off when you will lay your own track in an editor. You still pay the same credits. Audio is not a separate line item on this site.
A practical order of attempts
- Draft the prompt in a note, then paste it. The box holds up to 20,000 characters, which is more than a 30-second shot list needs.
- Run 5 or 8 seconds at 720P. Watch for extra people, on-screen text, and a camera move you did not ask for.
- Fix the prompt. Name the count of people. Say “no titles”. Describe the end pose.
- Only then move duration to 15 or 30, or resolution to 1080P. The 30-second guide covers that jump.
The seed field is optional. Leave it empty for a new roll. If a composition is close and you only want to nudge the prompt, set a seed and change one sentence. The same seed is not a guarantee of the same pixels, but it is the control the form gives you.
Sign-in is required. The MP4 link from the provider expires, so download the file if you want to keep it. Output from this site is cleared for commercial use under the terms on the generator page, which is a statement about our product, not a claim that every other Wan host works the same way.
Questions people hit on the first try
Why did the clip include a caption?
The prompt never forbade on-screen text, or it asked for a title. Say “no titles, no captions, no logos” in the same prompt. There is no separate negative box.
Can I upload a photo in text mode?
No. Switch to Image to Video for a first frame. Text mode ignores the idea of a still because the form does not send one.
Where are example prompts?
The prompt library at /wan-3-0-prompts. Each card has a Use this prompt link that prefills /wan-3-0.