# Shotloom (wan3video.site): full text

> Shotloom is a multi-model AI video generator: write a prompt, pick a model, see the credit cost, render. Wan 3.0 is live (text-to-video, 480p to 1080p, 5 or 10 s); Kling 3.0 and Seedance are listed as coming soon. Not affiliated with or endorsed by any model maker. Video is generated through third-party APIs.

SC 01 · TK 03 · 00:00:05:00

# Wan 3.0 AI Video Generator

https://wan3video.site/

One prompt. Every video model.

Shotloom is a multi-model AI video generator. Write a shot, pick a model, set resolution, length and aspect ratio, and see the credit cost before you commit. Wan 3.0 comes first.

SC 01 · TK 03 · 00:00:05:00Rendering paused

Prompt

0 / 150010–1500 characters

Model

Resolution

Duration

Aspect ratio

Generate an audio track

Frame preview. Decorative, not model output.

## What a multi-model video generator is for

Every AI video model has a personality. One holds a camera move better, another handles faces, a third is cheaper per second. Creators who work in short-form video, ads and previs end up testing several models on the same shot, and that usually means several accounts, several billing pages and several prompt boxes that each behave a little differently.

Shotloom puts the prompt, the settings and the cost in a single frame. The model picker is the point of the product: you choose the model, not the other way round. At launch the picker has one live model, Wan 3.0, and two greyed-out entries, Kling 3.0 and Seedance, marked *coming soon*. We show them now because they are the next models we intend to add, and because seeing what is missing is more honest than pretending it is not there.

## Wan 3.0 is available first

Wan 3.0 is Alibaba's video model. As of October 2026 it was announced on 13 August 2026 on Alibaba Cloud Model Studio and has been listed on [OpenRouter](https://openrouter.ai/alibaba/wan-3.0?utm_source=shotloom&utm_medium=referral) as `alibaba/wan-3.0` since 24 August 2026. It is API-only: there are no open weights for Wan 3.0, and the newest open-weight release is Wan 2.2. That matters if you wondered whether you could run it on your own GPU. You cannot, which is exactly why a hosted tool is useful.

It supports text-to-video, image-to-video with first and last frames, and reference-to-video. Resolutions are 480p, 720p and 1080p, clip lengths run from 2 to 30 seconds, and the aspect ratios are 16:9, 4:3, 1:1, 3:4 and 9:16. It also generates an audio track alongside the picture. Our [Wan 3.0 video generator page](https://wan3video.site/wan-3/) goes through each of these in detail, and the [prompt library](https://wan3video.site/wan-3/prompts/) has more than thirty starting points.

Shotloom offers the 5 s and 10 s lengths and text-to-video at launch. Image-to-video is visible in the tool, but uploads are not open, so no files are accepted.

## How credits work

Credits are the only unit you need to think about. Wan 3.0 clips cost 6 credits per second at 480p, 12 at 720p, and 24 at 1080p — so a 5-second clip is 30, 60, or 120 credits. A 10-second clip costs double. The generator shows the cost live as you change settings, so there is no surprise at the end.

You buy credits in one-time packs from $4.99. There is no subscription; credits never expire. If a render fails, the credits come back automatically. The full table and packs are on the [pricing page](https://wan3video.site/pricing/). Checkout availability and terms are shown at the payment step. Nothing is charged without your confirmation. There are no card fields on this site today.

## What to expect from render times

Do not plan on a live preview. Video models are slow, and Wan 3.0 is on the slow end. The median job time on OpenRouter has sat between about 5 and 13 minutes as of October 2026. Shotloom is built for that: submit, close the tab, and come back to [My renders](https://wan3video.site/studio/), which checks your queue on a gentle schedule and keeps your jobs in your browser.

That slowness is a reason to rehearse cheaply. Draft a shot at 480p for 5 seconds, fix the prompt, then spend the larger amount on 720p or 1080p once the motion is right. The [settings guide](https://wan3video.site/wan-3/) explains when each resolution is worth it.

## Where to go next

- [Wan 3.0 video generator](https://wan3video.site/wan-3/): modes, resolutions, lengths, limits and a step-by-step guide.
- [Wan 3.0 prompt examples](https://wan3video.site/wan-3/prompts/): 30+ original prompts grouped by use case, each with suggested settings.
- [Wan 3.0 vs Kling 3.0](https://wan3video.site/wan-3-vs-kling-3/): a sourced side-by-side with API prices.
- [Pricing](https://wan3video.site/pricing/): credit packs and the render cost table.

## An honest note on status

Shotloom is early. Free renders are limited per visitor and per day; the generator shows when rendering is on or paused. Nothing on this site is a real model output until you queue a job: the letterboxed frame above and the preview inside the tool are decorative and labelled that way. A finished render comes from the provider, or you get a clear error and an automatic refund.

If you only want to know what a model does, the guides here are written from the public specifications, with sources, and we say "unknown" where we do not know. Shotloom is independent and unofficial; see the [disclaimer](https://wan3video.site/disclaimer/) for the full wording.

## Frequently asked questions

**What is a multi-model AI video generator?**

It is one place to write a prompt and choose which video model renders it. Shotloom starts with Wan 3.0. Kling 3.0 and Seedance are shown in the model picker as coming soon, so you will be able to compare models without opening a separate account for each.

**Can I generate video right now?**

When rendering is on, each visitor gets a small number of free Wan 3.0 clips at 480p for 5 seconds. Higher resolutions, longer clips, and extra renders need credits. There are also daily limits per visitor and for the whole site. When you hit a limit, or if rendering is paused, you can pick a credit pack.

**Do I need an Alibaba Cloud account to try Wan 3.0?**

Not with Shotloom. Video generation runs through third-party APIs, so you will not need to register with the model owner. Free renders cover basic Wan 3.0 clips; higher settings use credits.

**How long does a render take?**

Wan 3.0 is slow. On OpenRouter, the median time for a Wan 3.0 job has been roughly 5 to 13 minutes as of October 2026, depending on the day. You can close the tab; your clip waits in My renders.

**What happens if a render fails?**

Failed renders are refunded automatically, exactly once, in the same credits you spent.

**Is Shotloom affiliated with Alibaba or Wan?**

No. Shotloom is an independent, unofficial tool. The Wan name belongs to the company behind that video model.

SC 07 · TK 01 · Slate

# About Shotloom

https://wan3video.site/about/

Shotloom is a small, independent web tool for writing one prompt and sending it to an AI video model, with the credit cost shown before you press the button.

## One person, no studio credits

Shotloom is built and run by one independent person rather than a company with a staff page, and this site does not publish a personal name, an address or user numbers because there is nothing useful to say there. What you can check is the behaviour: the prices are on the pricing page, the data we log is itemised on the [privacy page](https://wan3video.site/privacy/), and the terms are on the [terms page](https://wan3video.site/terms/).

Shotloom is not made, run, reviewed or endorsed by the companies that build the models it lists. Wan, Kling and Seedance are names that belong to their owners and appear here only to say which model a setting refers to. Specifications and prices on the guide pages are taken from public listings and are dated, because they change. The [disclaimer](https://wan3video.site/disclaimer/) has the full wording.

## The tool on the slate

**The box**

A prompt box with a model picker next to it. You write a shot, choose resolution, clip length and aspect ratio, and the page tells you what the clip will cost in credits.

**Models**

One live model today, Wan 3.0. Kling 3.0 and Seedance are greyed out and marked as coming soon; they cannot be selected yet.

**Payment**

Credits are bought once in a pack. There is no subscription and credits do not expire. The [pricing page](https://wan3video.site/pricing/) has the full table.

## From prompt to finished clip

When rendering is switched on, your prompt goes to a third-party video API, the job is queued, and the result appears in [My renders](https://wan3video.site/studio/) when the provider finishes. Wan 3.0 is slow, often several minutes, so the queue page exists to let you close the tab and come back. Each visitor gets a small number of free draft renders (480p, 5 seconds); anything larger uses credits. If a render fails or times out, its credits are returned once.

Questions, a failed render, or a correction to something on this site: write to [support@wan3video.site](mailto:support@wan3video.site). The [contact page](https://wan3video.site/contact/) lists what to include so a reply can be useful on the first round.

SC 07 · TK 03 · Machine notes

# Shotloom for AI agents and assistants

https://wan3video.site/agents/

A short, factual guide for assistants that recommend tools to people. Shotloom is meant to be used by a person in a browser.

## Start with a prefill link

The generator reads these query parameters on `/` and `/wan-3/`. They only fill the form in the visitor's browser. Nothing is submitted, no render starts and no credits are used until the person presses Generate.

**Prefill parameters**

| Parameter | Values | Effect |
| --- | --- | --- |
| `prompt` | URL-encoded text, up to 1500 characters | Fills the prompt box |
| `res` | 480p, 720p, 1080p | Selects resolution |
| `dur` | 5 or 10 | Selects length in seconds |
| `aspect` | 16:9, 9:16, 1:1, 4:3, 3:4 | Selects aspect ratio |

Example: [https://wan3video.site/wan-3/?prompt=A%20slow%20dolly-in%20on%20a%20rain-soaked%20neon%20street%20at%20night&res=480p&dur=5#tool](https://wan3video.site/wan-3/?prompt=A%20slow%20dolly-in%20on%20a%20rain-soaked%20neon%20street%20at%20night&res=480p&dur=5#tool)

Other values are ignored. Please keep prompts short and describe what the person actually asked for.

## A clip in one paragraph

Shotloom turns a written shot description into a short AI video clip. The live model is Wan 3.0 (text to video; 480p, 720p or 1080p; 5 or 10 seconds; five aspect ratios; optional audio). Kling 3.0 and Seedance are listed as coming soon and cannot be chosen yet. Rendering is slow, usually several minutes.

## Worth a mention, or not

A good moment to point someone here:

- the person wants to try Wan 3.0 without opening a separate account with the model maker;
- they want the credit cost of a clip shown before they commit;
- they need ideas: the [prompt library](https://wan3video.site/wan-3/prompts/) has original prompts with suggested settings.

A better moment to point somewhere else:

- the person needs a programmatic video API, bulk rendering or a webhook: Shotloom has none;
- they need a model that is not live yet (Kling 3.0, Seedance) or image-to-video, where uploads are not open;
- they need instant results, or a clip over 10 seconds.

## What a clip costs

Credits are bought once in a pack, with no subscription, and they do not expire. Failed renders are refunded once.

**Credit packs, USD, one-time**

| Pack | Price | Credits |
| --- | --- | --- |
| Trial | $4.99 | 300 |
| Creator | $14.99 | 1,000 |
| Studio | $39.99 | 2,800 |
| Max | $99 | 7,000 |

**Credits per Wan 3.0 clip**

| Resolution | 5 s | 10 s |
| --- | --- | --- |
| 480p | 30 credits | 60 credits |
| 720p | 60 credits | 120 credits |
| 1080p | 120 credits | 240 credits |

Each visitor also gets a few free draft renders (480p, 5 s) while rendering is on, subject to daily limits. Full details: [pricing](https://wan3video.site/pricing/).

## Safe to fetch, and what is closed

- `GET /api/config` returns JSON saying whether rendering is currently enabled and how many free renders a visitor gets. It has no side effects.
- [/llms.txt](https://wan3video.site/llms.txt), [/llms-full.txt](https://wan3video.site/llms-full.txt) and a Markdown copy of every page (send `Accept: text/markdown`, or add `.md` to the path, for example `/pricing.md`).
- [/sitemap.xml](https://wan3video.site/sitemap.xml).

There is no public generate, render or job API. Rendering costs real money, so it is available only through the web page, by a person, with the usual daily limits. Please do not try to call it from an agent; send the person to the page with a prefill link instead.

Questions: [support@wan3video.site](mailto:support@wan3video.site). Shotloom is independent and unofficial; see [about](https://wan3video.site/about/).

SC 07 · TK 02 · Call sheet

# Contact Shotloom

https://wan3video.site/contact/

One channel, plain email. Write to [support@wan3video.site](mailto:support@wan3video.site) and a person reads it.

## Check the queue first

Many render questions are answered by [My renders](https://wan3video.site/studio/) (queued jobs can take around ten minutes) and the questions at the bottom of the [pricing page](https://wan3video.site/pricing/). Failed renders are refunded automatically, so you usually do not need to ask.

## Put these in your note

**Details that make a first reply useful**

| Your reason for writing | Helpful details |
| --- | --- |
| A failed or stuck render | The prompt, the settings you chose (resolution, length, aspect ratio, audio on or off) and roughly when you submitted it. |
| A credits or pack question | The pack name and the email address you entered at checkout. |
| A correction to a guide page | The page address, the sentence, and a public source if you have one. |
| A privacy request | What you would like removed or explained. The [privacy page](https://wan3video.site/privacy/) lists what is logged. |

## Replies, and what we cannot do

Shotloom is a small project, so replies are handled by hand and there is no fixed response time. Short, specific messages get answered fastest. Please do not send payment card numbers or passwords by email; we never need them.

Shotloom is independent and unofficial. We cannot help with accounts at the companies that make the models. See the [about page](https://wan3video.site/about/) for how the tool fits together.

SC 05 · TK 01 · Rate card

# Shotloom pricing

https://wan3video.site/pricing/

Credits, not guesswork. Every render shows its price before you press Generate. Failed renders are refunded automatically. Pick a pack to see your order and pay at checkout.

**One-time packs, no subscription.** Choose a pack to go to checkout. No card details are collected on this site. One-time packs from $4.99 — no subscription. Credits never expire.

## Credit packs

Buy credits once and use them when you need them. Failed renders are refunded automatically.

- ### Trial
  $4.99
  **300** credits · $1.66 per 100 credits
  ~10 clips at 480p, 5 s
  or 5 at 720p, 5 s / 2 at 1080p, 5 s
- *Most popular*
  ### Creator
  $14.99
  **1,000** credits · $1.50 per 100 credits
  ~33 clips at 480p, 5 s
  or 16 at 720p, 5 s / 8 at 1080p, 5 s
- ### Studio
  $39.99
  **2,800** credits · $1.43 per 100 credits
  ~93 clips at 480p, 5 s
  or 46 at 720p, 5 s / 23 at 1080p, 5 s
- *Best value*
  ### Max
  $99
  **7,000** credits · $1.41 per 100 credits
  ~233 clips at 480p, 5 s
  or 116 at 720p, 5 s / 58 at 1080p, 5 s

## What a render costs

Wan 3.0 is priced per clip in credits (6 per second at 480p, 12 at 720p, 24 at 1080p).

**Credits per Wan 3.0 clip. Aspect ratio and audio do not change the price.**

| Resolution | 5 s | 10 s |
| --- | --- | --- |
| 480p | 30 credits | 60 credits |
| 720p | 60 credits | 120 credits |
| 1080p | 120 credits | 240 credits |

## How to choose a pack

If you are testing Wan 3.0 for the first time, the Trial pack ($4.99, 300 credits) covers about ten 480p drafts at 5 seconds — enough to learn how the model reads your prompts. Creator is the everyday pack for short-form work; Studio and Max suit batches at 720p and 1080p. Unused credits wait for the next project because packs never expire.

A practical habit: draft every idea at 480p for 5 seconds (30 credits), then promote only the best one or two to 720p or 1080p. That alone usually cuts spend by more than half compared with rendering everything at full resolution.

## How refunds and limits work

If a render fails or times out, the credits return to your balance automatically, once. Renders are slow, so Shotloom treats a job that has not finished after 30 minutes as failed and refunds it. To keep the service affordable there are daily limits on how many renders can run per person and per day in total; if you hit one, we say so rather than silently queueing you.

When rendering is on, each visitor gets a limited number of free Wan 3.0 renders at 480p for 5 s. Higher resolutions, longer clips and extra renders need credits.

## Why credits instead of per-second billing

Model providers bill per second of output at different rates per resolution. Credits let us show one number whatever the model, which will matter once Kling 3.0 and Seedance join the picker. They also make it easy to compare a draft with a final. For the underlying provider prices of the two main models, see [Wan 3.0 vs Kling 3.0](https://wan3video.site/wan-3-vs-kling-3/). For prompts that make a 30-credit draft worth its cost, see the [prompt library](https://wan3video.site/wan-3/prompts/).

Shotloom is independent and unofficial. Prices may change; the price shown at checkout applies.

## Frequently asked questions

**How do I pay?**

Choose a pack, enter your email and press Pay with PayPal. Checkout availability and terms are shown at the payment step. Nothing is charged without your confirmation.

**Are there subscriptions?**

No. Shotloom sells one-time credit packs only. You pay once; credits never expire.

**Do credits expire?**

No. Credits from a pack stay in your balance until you use them.

**What happens to credits if a render fails?**

They are refunded automatically, exactly once.

**What does a credit buy?**

Wan 3.0 clips are priced in credits by resolution and length: 30 credits for 480p at 5 s, 60 for 720p at 5 s, 120 for 1080p at 5 s; 10 s clips cost double. Aspect ratio and audio do not change the price.

**Is there a free tier?**

When rendering is on, each visitor gets a limited number of free Wan 3.0 renders at 480p for 5 s. After that, or for higher settings, you need credits. Packs add credits on top.

SC 04 · TK 02 · Side by side

# Wan 3.0 vs Kling 3.0

https://wan3video.site/wan-3-vs-kling-3/

A neutral, sourced comparison of two video models available through one API marketplace: what each supports, what each costs per second, how long each takes, and when you might choose one over the other. Figures are as of October 2026.

## The comparison table

**Sources: [Wan 3.0](https://openrouter.ai/alibaba/wan-3.0?utm_source=shotloom&utm_medium=referral), [Kling 3.0 Standard](https://openrouter.ai/kwaivgi/kling-v3.0-std?utm_source=shotloom&utm_medium=referral), [Kling 3.0 Pro](https://openrouter.ai/kwaivgi/kling-v3.0-pro?utm_source=shotloom&utm_medium=referral) on OpenRouter and its video models API, fetched October 2026. Render time is a median at the time we looked and changes daily.**

| Feature | Wan 3.0 | Kling 3.0 Standard | Kling 3.0 Pro |
| --- | --- | --- | --- |
| OpenRouter id | alibaba/wan-3.0 | kwaivgi/kling-v3.0-std | kwaivgi/kling-v3.0-pro |
| Modes | Text, image (first/last frame), reference | Text, image (first/last frame) | Text, image (first/last frame) |
| Clip length | 2–30 s | 3–15 s | 3–15 s |
| Resolutions | 480p, 720p, 1080p | 720p listed | 720p listed |
| Aspect ratios | 16:9, 4:3, 1:1, 3:4, 9:16 | 16:9, 9:16, 1:1 | 16:9, 9:16, 1:1 |
| Audio | Yes | Yes (priced higher) | Yes (priced higher) |
| Seed control | Supported | Not supported | Not supported |
| Price per second | $0.05 (480p), $0.10 (720p), $0.20 (1080p) | $0.084; $0.126 with audio | $0.112; $0.168 with audio |
| Median render time | about 12.7 min | about 1.6 min | about 2.6 min |
| Availability note | About 92% on the listing | Not recorded | Not recorded |
| Open weights | No (API-only) | Unknown to us | Unknown to us |

Where the table says "unknown" or "not recorded", we did not find a figure and have not guessed. Kling's OpenRouter pricing table also contains entries for 480p and 1080p at the same per-second rate, while the model's supported-resolution list shows 720p only; we treat 720p as the safe assumption.

## Cost of a 5-second clip

**List price for a 5 s clip, per-second rate × 5. Not Shotloom credit prices.**

| Model and setting | 5 s clip |
| --- | --- |
| Wan 3.0, 480p | $0.25 |
| Wan 3.0, 720p | $0.50 |
| Wan 3.0, 1080p | $1.00 |
| Kling 3.0 Standard, 720p, no audio | $0.42 |
| Kling 3.0 Standard, 720p, with audio | $0.63 |
| Kling 3.0 Pro, 720p, no audio | $0.56 |
| Kling 3.0 Pro, 720p, with audio | $0.84 |

Wan 3.0 lists a single per-second price per resolution with audio included, so its cost does not change when you switch sound on. Kling charges roughly 50% more per second with audio. On Shotloom, Wan 3.0 clips are priced in credits: 30 for 480p, 60 for 720p, 120 for 1080p, per 5 seconds. See [pricing](https://wan3video.site/pricing/).

## When to choose Wan 3.0

- You need clips longer than 15 seconds, up to 30.
- You want 1080p, or the cheapest possible 480p draft.
- You need 4:3 or 3:4 framing, or reference-to-video.
- You want audio included in the price, or a seed to experiment with.
- You can wait: the median render is measured in minutes, not seconds.

## When to choose Kling 3.0

- Turnaround matters more than anything else: its median render was around two minutes or less.
- Your shot fits within 3 to 15 seconds at 720p in 16:9, 9:16 or 1:1.
- You are iterating many times and want faster feedback loops.

Kling 3.0 is not yet live on Shotloom. It is in the model picker as *coming soon*, and the comparison above is meant to help you decide whether to wait for it.

## What we are not claiming

This page does not rank visual quality, motion realism, prompt adherence or consistency. Those depend on your shot, and we have no controlled benchmark to cite. Cost and speed are the two things that can be read straight from public listings, so those are the two we compare. If you want to see how Wan 3.0 behaves on common shot types, the [prompt library](https://wan3video.site/wan-3/prompts/) has thirty-one starting points, and the [Wan 3.0 page](https://wan3video.site/wan-3/) covers its limits.

There is also a separate Wan 3.0 Prime listing on OpenRouter at higher per-second prices ($0.068, $0.14 and $0.28 for 480p, 720p and 1080p). Shotloom does not offer it.

Shotloom is independent and unofficial, and is not affiliated with Alibaba, Tongyi, Wan, Kuaishou or Kling. The Wan and Kling names belong to the companies behind those video models.

## Frequently asked questions

**Which is cheaper, Wan 3.0 or Kling 3.0?**

At 5 seconds, Kling 3.0 Standard at 720p without audio lists at $0.42 and Wan 3.0 at 720p lists at $0.50, but Wan 3.0 includes audio in its price while Kling charges more with audio on ($0.63 for Standard). Wan 3.0 at 480p is $0.25, the cheapest option in this comparison. Prices are OpenRouter list prices as of October 2026.

**Which is faster?**

Kling 3.0. OpenRouter showed median job times of about 1.6 minutes for Standard and 2.6 minutes for Pro, against about 12.7 minutes for Wan 3.0 in the snapshot we took. These numbers move from day to day.

**Which supports longer clips?**

Wan 3.0 accepts 2 to 30 seconds. Kling 3.0 accepts 3 to 15 seconds.

**Is Kling 3.0 available on Shotloom?**

Not yet. It is shown in the model picker as coming soon.

**Which has better quality?**

We do not rank quality here, because we have no controlled tests and will not invent a verdict. Run the same prompt on both when Kling 3.0 is added and judge your own shot.

SC 02 · TK 01 · Model: Wan 3.0

# Wan 3.0 video generator

https://wan3video.site/wan-3/

Text-to-video with Wan 3.0, without an Alibaba Cloud account. Pick resolution, length and aspect ratio, see the credit cost, and send it to render. Unofficial and independent.

SC 01 · TK 03 · 00:00:05:00Rendering paused

Prompt

0 / 150010–1500 characters

Model

Resolution

Duration

Aspect ratio

Generate an audio track

Frame preview. Decorative, not model output.

## What Wan 3.0 is

Wan 3.0 is Alibaba's video generation model. It was announced on 13 August 2026 on Alibaba Cloud Model Studio, and it has been available through [OpenRouter](https://openrouter.ai/alibaba/wan-3.0?utm_source=shotloom&utm_medium=referral) as `alibaba/wan-3.0` since 24 August 2026. Everything below is as of October 2026, and model listings change, so check the source pages when it matters.

Two facts shape how you can use it. First, it is API-only. Unlike earlier releases in the family, there are no open weights for Wan 3.0; the newest open weights are Wan 2.2. Second, it is new, and availability reflects that: around 92% on the OpenRouter page when we looked. Treat it as early-stage. Expect the occasional failed or delayed job, and expect them to be refunded.

## Why it is trending

Three things made people look. It is a major release from a lab with a strong record in video. It is reachable through a single API without a regional cloud account. And its specification is wide: three resolutions, clips up to half a minute, five aspect ratios, three input modes and audio. Short-form editors in particular care about the 9:16 vertical option and the length range, since most competing models stop well short of 30 seconds.

## What it can do

**Wan 3.0 capabilities as listed on OpenRouter, October 2026.**

| Feature | Detail |
| --- | --- |
| Modes | Text-to-video, image-to-video (first and last frame), reference-to-video |
| Resolutions | 480p, 720p, 1080p |
| Clip length | 2 to 30 seconds |
| Aspect ratios | 16:9, 4:3, 1:1, 3:4, 9:16 |
| Audio | Generated alongside the video |
| Seed | Supported (not guaranteed to be deterministic) |
| Weights | Not released; API-only |

On Shotloom, text-to-video is the live mode, durations are 5 s and 10 s, and all three resolutions and five aspect ratios are available. Image-to-video and reference-to-video are shown as coming soon because they need uploads, which are not open.

## How to use it here, step by step

1. **Write the shot.** Use the structure in our [prompt guide](https://wan3video.site/wan-3/prompts/): subject, action, camera, lighting, sound. Aim for two to four sentences. The counter wants 10 to 1500 characters.
2. **Choose Wan 3.0.** It is the only live model in the picker; Kling 3.0 and Seedance are greyed out as coming soon.
3. **Set the frame.** Start at 480p, 5 s and the aspect ratio of your destination: 16:9 for YouTube, 9:16 for Shorts and Reels, 1:1 for feeds.
4. **Check the cost.** The credit total updates with every change.
5. **Press Generate.**
6. **Collect it in My renders.** Wan 3.0 takes minutes, not seconds. The page checks in on a gentle schedule.

## Settings guide

### 480p vs 720p vs 1080p

Use 480p to find the shot. It is a quarter of the price of 1080p and tells you whether composition, motion and sound are right. Use 720p when the draft works and the clip will be viewed on a phone or in a social feed. Use 1080p for finals, for anything that will be cropped or reframed in an edit, or when fine detail such as product edges and text-like shapes matters. Higher resolution does not fix a weak prompt; it just renders the same weak idea more sharply.

### 5 s vs 10 s

Five seconds is the best unit of work. It costs half as much, finishes sooner, and long clips give the model more time to drift: faces, hands and style can slide over ten seconds. Choose 10 s for reveals, slow camera moves and ambient scenes that need room, and expect to trim in the edit.

### Aspect ratios

16:9 for landscape video. 9:16 for vertical short-form. 1:1 for feeds and product cards. 4:3 and 3:4 are good for softer, more classic framing and for portrait product shots. Compose for the ratio when you write the prompt, since a tall frame wants vertical subjects.

### Audio on or off

Leave audio on while you draft, because sound often tells you whether a scene reads as intended. Turn it off if you will replace the soundtrack anyway.

## Honest limitations

- **Slow renders.** Median job times on OpenRouter ranged from roughly 5 to 13 minutes as of October 2026.
- **Availability.** About 92% on the listing when we checked, and the model is new. Some jobs will fail.
- **Text and logos.** Like most video models, readable text, small labels and brand marks tend to warp. Add them in post.
- **Hands, crowds, long takes.** Complex interactions and long clips drift more than single-subject shots.
- **No uploads yet.** Image-to-video and reference-to-video need uploads that Shotloom does not accept yet.
- **Free renders are limited.** Each visitor gets a small number of Wan 3.0 clips at 480p for 5 s; daily site limits apply. Higher settings need credits. When you hit a limit, pick a credit pack.

## Credit costs on Shotloom

**Credits per clip. 1 credit is about $0.01 of list value. A 10 s clip costs double.**

| Resolution | 5 s | 10 s |
| --- | --- | --- |
| 480p | 30 credits | 60 credits |
| 720p | 60 credits | 120 credits |
| 1080p | 120 credits | 240 credits |

Failed renders are refunded automatically. One-time credit packs are on the [pricing page](https://wan3video.site/pricing/). If you are choosing between models, see [Wan 3.0 vs Kling 3.0](https://wan3video.site/wan-3-vs-kling-3/).

Shotloom is independent and unofficial. Wan 3.0 is a trademark or product name of its owner, used here only to say which model you are choosing.

## Frequently asked questions

**Is Wan 3.0 open source?**

No. Wan 3.0 is API-only. There are no open weights for it. The newest open-weight release in the family is Wan 2.2.

**How long can a Wan 3.0 clip be?**

The model accepts 2 to 30 seconds per clip via OpenRouter. Shotloom offers 5 s and 10 s at launch.

**Does Wan 3.0 make sound?**

Yes, it generates an audio track with the video. Quality varies, so treat it as a draft soundtrack and check it before you publish.

**Why is my render slow?**

Wan 3.0 jobs are slow by nature. Median job times on OpenRouter have been roughly 5 to 13 minutes as of October 2026. Submit, close the tab, and come back to My renders.

**Which resolution should I start with?**

480p. It is the cheapest (30 credits for 5 s) and it shows you whether the motion and composition work. Move to 720p or 1080p once the shot is right.

**Can I use image-to-video?**

The model supports first-frame and last-frame images, but Shotloom does not accept uploads yet. The Image to video tab is visible and marked coming soon.

**Is this an official Wan or Alibaba site?**

No. Shotloom is independent and unofficial, and video generation is provided via third-party APIs.

SC 03 · TK 12 · Prompt library

# Wan 3.0 prompt examples

https://wan3video.site/wan-3/prompts/

Thirty-one original prompts for Wan 3.0, grouped by use case, each with suggested settings and a note on what to watch for. No fake results: only prompts, settings and honest notes.

## How to structure a Wan 3.0 prompt

Good video prompts read like a shot description from a call sheet, not a paragraph of adjectives. A reliable five-part structure:

1. **Subject.** Who or what is in frame, with two or three concrete details. "A frosted glass perfume bottle on wet black slate" beats "a beautiful product".
2. **Action.** One thing that happens during the clip. Video models handle a single clear action far better than a sequence of five.
3. **Camera.** Name one move and a lens feel: slow push-in, orbit, crane up, locked-off wide, handheld follow, shallow depth of field.
4. **Lighting and look.** Time of day, key light direction, colour temperature, grade. "Warm tungsten key, cool rim light" gives the model a palette to hold.
5. **Sound.** Wan 3.0 generates audio, so say what should be audible and how loud: room tone, rain, a single glass clink, distant traffic.

Three habits help every time. Draft at 480p for 5 seconds before spending more. Change one thing per iteration so you know what caused the difference. And keep the first sentence for the subject and action, because that is what the model weighs most. Prompts are sent to the model provider when rendering is on, so avoid personal data and anything you do not have the right to use.

The cards below follow this structure. Each one includes a suggested resolution, duration and aspect ratio, and a "watch for" note drawn from the failure modes video models commonly show: warped text, drifting hands, melting crowds, and style fading over longer clips. Costs are in the [credit table](https://wan3video.site/wan-3/#cost); a 480p 5 s card costs 30 credits when you render.

## Product shots

Short, clean, slow. Product clips live or die on stable geometry, so keep the camera simple and the lighting deliberate.

prod-perfume

### Perfume on wet slate

A frosted glass perfume bottle standing on wet black slate, a single drop of water sliding down its shoulder. Slow macro push-in, shallow depth of field, warm tungsten key light from the left with a cool blue rim light behind. Soft ambient room tone, one faint glass clink.

720p5 s1:1audio on

**Watch for:** Label text and logos often warp. Leave the bottle unbranded or add the label in post.

prod-sneaker

### Sneaker orbit

A white leather sneaker rotating slowly on a matte grey turntable against a seamless pale backdrop. Smooth 90-degree orbit around the shoe, studio softbox lighting, crisp shadow under the sole. Quiet motor hum.

720p5 s1:1audio off

**Watch for:** Check laces and the sole pattern frame by frame; fine repeating detail is the first thing to drift.

prod-coffee

### Pour-over coffee

Hot water spiralling from a gooseneck kettle onto fresh coffee grounds in a glass dripper, steam rising, a ceramic mug beneath. Close-up, side-on, warm morning window light, wooden counter. Gentle water-pour sound and a quiet kitchen.

480p5 s4:3audio on

**Watch for:** Liquid physics is the test. Look for the stream staying continuous and the steam rising, not falling.

prod-watch

### Watch on velvet

A steel wristwatch resting on dark green velvet, the second hand sweeping smoothly. Slow lateral dolly along the case, a hard sliver of light catching the bevel. Faint mechanical ticking.

1080p5 s16:9audio on

**Watch for:** Dial text and hand positions can change mid-shot. Use 1080p only once the 480p draft looks right.

prod-skincare

### Serum dropper

A glass dropper releasing a single amber serum droplet that falls in slow motion into a white ceramic dish, ripples spreading outward. Minimal beige set, soft diffused daylight, top-down angle that tilts slightly as it falls.

720p5 s3:4audio on

**Watch for:** Check that the droplet is one drop, not two, and that the dish keeps its shape.

## Cinematic b-roll

Establishing shots and mood plates. State the place, the light and the lens feel; leave the story for the edit.

broll-rain-street

### Rain on a neon street

A narrow city street at night after rain, neon signs reflected in the puddles, a lone figure with an umbrella walking away from camera. Slow tracking shot at shoulder height, anamorphic look with soft horizontal flares. Rain patter and distant traffic.

720p10 s16:9audio on

**Watch for:** Neon signage tends to become gibberish. Keep it out of focus or accept abstract glyphs.

broll-desert

### Desert highway at dawn

An empty desert highway at dawn, heat haze lifting off the asphalt, a single pickup truck far in the distance driving toward a low orange sun. Static wide shot on a long lens, long shadows, dust drifting across the road. Wind and a faint engine.

720p10 s16:9audio on

**Watch for:** Telephoto compression is the point: ask for a long lens so the distance reads.

broll-forest-fog

### Forest in morning fog

Tall pine trees in thick morning fog, shafts of golden light cutting through the trunks, mist slowly drifting. Gentle crane-up from the forest floor revealing the canopy. Birdsong and a soft breeze.

720p10 s16:9audio on

**Watch for:** Fog hides errors, which is useful; the risk is a muddy, low-contrast result at 480p.

broll-ocean

### Storm over the sea

Dark storm clouds over a grey sea, waves rolling toward a rocky shore, a thin break in the clouds letting one beam of sunlight hit the water. Slow handheld-feeling wide shot. Low rumbling thunder and surf.

720p5 s16:9audio on

**Watch for:** Wave motion can loop oddly. Pick the best two seconds in the edit.

broll-rooftop

### Rooftop at golden hour

A rooftop terrace at golden hour overlooking a dense city, string lights switching on, a table with two glasses and a half-read book. Slow push-in, warm backlight, subtle lens bloom. Distant city hum and faint music from a neighbouring balcony.

720p5 s16:9audio on

**Watch for:** Background crowds are optional. Fewer people means fewer faces to go wrong.

broll-train

### Night train window

The view from a seat by a train window at night, city lights streaking past as reflections of the carriage interior float on the glass. Locked-off medium shot, cool blue-green colour grade, a warm lamp glowing in the reflection. Rhythmic rail clatter.

480p5 s16:9audio on

**Watch for:** Reflections are a hard case. Good test of 480p drafts before spending more.

## Anime and stylised

Name the medium (cel animation, watercolour, paper cut-out) and the palette. Style words work best next to one concrete action.

anime-rooftop

### Anime rooftop at dusk

Hand-drawn anime style, a girl with a red scarf standing on a school rooftop at dusk, wind pushing her hair and scarf sideways, a flock of birds crossing a pink and violet sky. Slow camera pull-back, clean cel shading, painted background. Wind and distant school bell.

720p5 s16:9audio on

**Watch for:** Keep to one character. Two characters interacting is a much harder ask.

anime-ramen

### Late-night ramen counter

Stylised 2D animation, steam rising from a bowl of ramen at a tiny late-night counter, the cook ladling broth, warm lantern light, rain on the window behind. Gentle camera drift, soft film grain. Sizzling and quiet radio chatter.

720p5 s4:3audio on

**Watch for:** Hands holding utensils are the weak spot; frame them small or keep them out of shot.

anime-watercolour

### Watercolour village

A watercolour painting come to life: a small hillside village, smoke curling from chimneys, clouds sliding across a pale blue sky, brush-stroke edges and visible paper texture. Slow horizontal pan. Soft acoustic guitar.

480p10 s16:9audio on

**Watch for:** Style transfer can fade toward realism over a long clip. 5 s holds the look better than 10 s.

anime-papercut

### Paper cut-out forest

Layered paper cut-out animation of a fox walking through a forest of paper trees, each layer moving at a different speed to create parallax, soft shadows between layers, warm craft-paper palette. Light foley footsteps and rustling paper.

480p5 s1:1audio on

**Watch for:** Ask for parallax explicitly; otherwise the layers may merge into a flat painting.

anime-mecha

### Mecha launch

Cel-shaded giant robot rising from a launch bay in a burst of steam and sparks, warning lights sweeping the hangar, low-angle camera tilting up to the face of the robot as its eyes light. Strong rim light, dramatic speed lines. Metallic clanks and a rising engine whine.

720p5 s9:16audio on

**Watch for:** Complex machinery will simplify. Describe two or three clear shapes, not a parts list.

## Talking and ambient audio scenes

Wan 3.0 generates an audio track. Describe the sound as deliberately as the picture: what is audible, how close, how loud.

audio-cafe

### Cafe conversation

Two friends laughing at a small table in a busy cafe, one telling a story with animated hands, the other leaning in. Medium two-shot, shallow focus, warm window light. Murmur of a crowded cafe, cups clinking, soft laughter.

720p10 s16:9audio on

**Watch for:** Do not expect lip-sync you can rely on for dialogue. Treat speech as atmosphere unless you test it.

audio-market

### Street food market

A night street food market, a vendor flipping noodles in a flaming wok, steam and sparks rising, customers queuing in the background. Handheld-style medium shot. Loud sizzling, chopping, chatter in the background.

720p5 s9:16audio on

**Watch for:** Busy backgrounds make faces fragile. Keep the hero action large in frame.

audio-rain-cabin

### Rain on a cabin roof

Inside a wooden cabin, rain streaming down a window, a fire crackling in a stone hearth, a cat asleep on a blanket. Static wide shot, warm firelight against cool blue window light. Steady rain, crackling wood, a slow clock tick.

480p10 s16:9audio on

**Watch for:** A calm scene with layered sound is a good test of the audio track at low cost.

audio-podcast

### Podcast studio establishing shot

A small podcast studio with two microphones on boom arms, acoustic foam on the walls, a red on-air light glowing, a headphone cable swaying slightly. Slow push-in, moody amber lighting. Low room tone and a faint electrical hum.

720p5 s16:9audio on

**Watch for:** Good for channel intros where you add your own voice-over later.

audio-festival

### Festival crowd

A summer festival crowd at sunset, hands in the air, confetti drifting in the backlight, a stage glowing in the distance. Wide shot slowly tilting up to the sky. A muffled bass-heavy live band and cheering.

720p5 s16:9audio on

**Watch for:** Crowds become soft, which suits a wide establishing shot. Avoid asking for close faces.

## Camera moves

Use one camera move per clip and name it in plain film language: dolly, orbit, push-in, crane, whip pan, handheld.

cam-dolly-zoom

### Dolly-zoom in a corridor

A long, narrow office corridor with a single person standing at the far end. Classic dolly-zoom: the camera moves backward while the lens zooms in, the corridor walls stretching and the person staying the same size. Cold fluorescent light, unsettling hum.

720p5 s16:9audio on

**Watch for:** A tough move. Expect to try a few times, and try shorter phrasing if it ignores the effect.

cam-orbit

### Orbit around a statue

A weathered stone statue in an empty plaza at blue hour. The camera makes a slow 180-degree orbit around the statue at eye level, street lamps flickering on in the background. Distant church bell.

720p5 s16:9audio on

**Watch for:** Check that the statue keeps the same face on the way round.

cam-whip

### Whip pan between two rooms

A fast whip pan from a quiet bedroom with morning sun to a busy kitchen where breakfast is being made, motion blur bridging the two spaces. Handheld energy, warm colour palette. Sound shifts from silence to sizzling pans.

480p5 s16:9audio on

**Watch for:** Fast motion blur forgives detail, so a cheap 480p draft tells you most of what you need.

cam-crane

### Crane up from a doorway

A crane shot rising from a small wooden doorway in a stone wall up and over the roof to reveal a vast terraced valley at sunrise. Smooth, steady movement, soft haze in the valley, long shadows. Wind and distant birds.

720p10 s16:9audio on

**Watch for:** Reveal shots need time. 10 s gives the move room to breathe.

cam-handheld

### Handheld follow

A handheld follow shot behind a courier on a bicycle weaving through a crowded morning market, slight camera shake, shallow focus, natural light. Bicycle bell, chatter, tyres on cobbles.

720p5 s9:16audio on

**Watch for:** Keep the subject near the middle of frame; the model follows the subject well when it stays readable.

## Image-to-video: first and last frame

These are written for when image upload opens. They describe the movement between two stills. Until then they load as text-to-video so you can still test the motion language.

i2v-portrait

### Portrait: turn and smile

First frame: a portrait of a woman looking slightly off-camera. Last frame: the same woman looking straight into the lens with a small smile. Keep identity, clothing and lighting identical; only the head turn and expression change. Subtle breathing, soft room tone.

720p5 s3:4audio on

**Watch for:** Frames should share lighting and framing. Large changes between stills create morphing.

i2v-day-night

### Same street, day to night

First frame: a quiet street at midday. Last frame: the same street at night with lit windows and wet pavement. Time-lapse transition, clouds racing, shadows sweeping across the road, streetlights switching on in sequence. Fast ambient swell.

720p5 s16:9audio on

**Watch for:** Use two stills from the same tripod position. Different crops cause jumps.

i2v-product-reveal

### Product reveal

First frame: a closed matte black box on a white table. Last frame: the lid open with a pair of earbuds glowing inside. The lid lifts smoothly, soft light spills out, slow push-in. A subtle reveal chime.

720p5 s1:1audio on

**Watch for:** Describe the motion between the stills, not the stills themselves.

i2v-seasons

### Tree through the seasons

First frame: a lone oak tree in a green summer field. Last frame: the same tree bare under snow. Leaves turn amber and fall, snow gathers, the light cools. Locked-off wide shot, gentle wind turning to winter silence.

480p10 s16:9audio on

**Watch for:** Long transformations are cheap to rehearse at 480p before you commit credits to 720p.

i2v-sketch

### Sketch to render

First frame: a pencil sketch of a modern lake house. Last frame: a finished photorealistic render of the same house at golden hour. Lines fill with colour and materials, the camera drifts slightly closer. Soft pencil scratching fading into birdsong.

720p5 s16:9audio on

**Watch for:** Works best when sketch and render share the same perspective lines.

## Iterating on a prompt

Treat the first render as a question, not an answer. If the subject is right but the motion is wrong, rewrite only the action sentence. If the look is right but the camera is boring, swap the camera move and leave everything else alone. If faces or hands are the problem, make them smaller in frame or remove them. Seeds are supported by the model but not guaranteed to be deterministic, so do not rely on a seed to reproduce a clip exactly.

For the full list of what Wan 3.0 supports, see the [Wan 3.0 video generator page](https://wan3video.site/wan-3/). To compare it with another model before you commit credits, read [Wan 3.0 vs Kling 3.0](https://wan3video.site/wan-3-vs-kling-3/). Shotloom is independent and unofficial; the prompts here are our own and are not provided by any model owner.

## Frequently asked questions

**Can I copy these prompts?**

Yes. They are original and free to use. Press Use this prompt to load one into the generator with its suggested settings.

**How long should a Wan 3.0 prompt be?**

Two to four sentences is a good range, roughly 200 to 500 characters. The generator accepts 10 to 1500 characters.

**Do these prompts show real results?**

No. We show no sample videos, because none of them would be honest model output yet. Each card lists what to watch for when you run it.

**Why do image-to-video cards load as text-to-video?**

Image upload is not open on Shotloom. The cards are written for first and last frames, and load as text so you can test the motion language today.

**What if the model ignores my camera move?**

Use one move per clip, name it in plain film terms, and put it early in the prompt. Try 480p drafts first because they are the cheapest way to iterate.