TJP Quickstart: From Nothing to Your First AI Video Clip
This is one complete video, start to finish, built from scratch — you describe what you want and the tool makes it.
If you are a complete beginner, this is not the fastest route. Starting from a look that already works, swapping in your own character and changing it takes a few clicks and no prompt writing at all. That path is copy, change, publish, and most people should start there.
Come back here when you need something that does not exist yet — which is the one thing copying cannot give you.
Follow it exactly the first time and experiment afterwards; the point of the first run is to see the whole pipeline work, not to make something good.
Budget about an hour. Most of that is waiting.
Data Notice: Tool names, model names and panel layouts in AI media platforms change frequently. The sequence below reflects the platform at time of writing; if a control has moved, the underlying logic is unchanged.
Before you start
Create an account at tjp.com. The free plan is a real plan rather than an expiring trial — it generates on a pay-as-you-go basis, which is enough to complete everything on this page.
You do not need a camera, editing software, drawing ability, or any idea of how the technology works.
1. Open the workspace and look at the shape of it
Sign in and open the studio. The layout is the same throughout:
- Left — a menu of everything you can make.
- Middle — the tool you picked, and its controls.
- Right — your library. Empty for now; everything you make lands here.
Nothing loads until you choose a tool, which is why the middle looks bare on arrival. There are also guided options in the middle — a tour and an in-app walkthrough — covering the same ground from inside the product.
2. Do not start with the flagship model
The best image model on the platform, Apex Image, does not run until you have an avatar connected. That is a detour, and it is the single most common place a first session stalls.
For a first run, pick either Apex img2img if you have a photo to work from, or one of the Partner Models listed lower in the menu, which need nothing set up.
The controls change depending on which tool you pick. That is normal — not every model accepts every setting.
3. Make a still first, not a video
This is the habit that matters most, and it is worth understanding why before you do it.
Video costs many times what a still costs and takes minutes rather than seconds. A still answers nearly every question you have — is the framing right, is the wardrobe right, is the light right — for a fraction of the price. Settling those on stills and only then animating the approved frame is the difference between an affordable production and an expensive one.
Type this, exactly:
a woman in a red coat standing at a rainy bus stop at night,
neon shop signs behind her, wet reflections on the pavement,
cinematic film still, shallow depth of field
Four things are doing work there, and it is worth noticing which:
| Part | What it sets | In the example |
|---|---|---|
| Subject | who or what, doing what | a woman in a red coat, standing |
| Setting | where | a rainy bus stop at night, neon signs |
| Light | how it is lit | neon, wet reflections |
| Treatment | what kind of picture this is | cinematic film still, shallow depth of field |
Subject alone gives you something generic. Subject plus setting plus light plus treatment is what makes an image look deliberate. When a result comes out flat, the missing part is almost always light or treatment.
4. Leave the settings alone
There will be controls for size, quantity, quality, seed and more. Change none of them yet. The defaults are chosen to work, and changing several at once means that when a result is wrong you will not know which change did it.
One exception worth making: if there is a how many control, set it to four. Each costs the same as one, so four images costs four times one image — but four gives you a choice where one gives you a verdict. While you are still learning what a prompt does, that is the best value on the panel.
Set the shape to portrait if you are making anything for a phone. Getting this right at the start beats cropping later, because cropping throws away part of the picture.
5. Generate
Press generate. The cost is shown on the button before you commit — you never find out afterwards.
Your images appear in the library area, first as placeholders and then finished. A still usually takes seconds to a minute. If a lot of people are generating at once, yours waits its turn and the status says where it is.
You can leave the page. The work carries on without you and the results are in your library when you come back.
6. Pick one and change exactly one word
Open the four full size and choose the best one.
Now go back to the prompt — it is still there — and change one word. Make the coat blue, or the night into dawn. Generate again.
Comparing two images that differ in one word teaches you more than ten images that differ in everything. This is the habit that separates people who get what they wanted from people who keep rolling the dice.
7. Now animate the frame you kept
Choose Apex Video from the menu under Create. The panel changes, because video needs settings a still does not.
There are two ways in, and they are not equally good:
- From a description. Write what should happen. Fast, and much less predictable.
- From a picture. Supply the still you already approved and describe how it should move. This is the better option most of the time, because the composition is already settled — you are only directing motion.
Use the second. Supply your kept frame, and write the motion:
she turns her head slowly toward the camera, rain continues to fall,
camera holds steady
Two kinds of motion both need saying out loud:
- Subject motion — what the things in the shot do.
- Camera motion — push in, pan left, or nothing.
If you do not mention the camera you get whatever the model decides, which is usually more movement than you wanted. “Camera holds steady” is one of the most useful phrases available to you.
Ask for one clear action. A clip a few seconds long cannot deliver a sequence of events, and asking for three things usually produces none of them cleanly.
8. Wait properly
Video takes minutes rather than seconds, and the cost is again shown before you commit.
You do not need to sit on the page. The work continues after you navigate away and the finished clip is in your library when you return. Turn on notifications if you would rather be told.
9. Download it
Open the clip and download the original file. Downloading your own work is free and unlimited — only generating costs anything.
That file is yours. It is also, at this point, a shot rather than a video: seconds long, one action, no sound. Turning shots into something people watch is the subject of from clips to a finished episode.
What just happened, and what it cost
You ran the entire production pipeline: describe, choose, animate, export. The expensive parts of that — the crew, the location, the camera package, the talent — did not exist, and the parts that remained took under an hour.
That is the compression this whole category is about. What it does not include is an audience or a payment, which run on a much slower clock. See where the money actually comes from for what that actually looks like.
What to read next
- Writing a short-drama script an AI tool can shoot — because the next bottleneck is knowing what to generate.
- Same face every episode — the feature that turns clips into a series.
- Prompting for video — motion, camera and light in more depth.
- What an episode actually costs — before you scale any of this up.
Last reviewed: · Editorial policy · Report an error