Your first 30 minutes with Deevid AI

If you’ve just opened a Deevid AI account, congratulations — you now have a platform that can do a remarkable amount. The problem: the first 30 minutes can go one of two ways. Structured, you leave with a publishable clip and a clear mental model. Unstructured, you leave with a folder of mediocre renders and the vague sense that you’ve “tried it.”

This walkthrough is the structured version.

Pick a goal before you write a prompt

Open Deevid and resist the urge to start prompting. First, decide what kind of output you want. The tool has presets that tune the prompt enhancer and the default render settings toward a specific category — picking the right one at the start saves you 20 minutes of fighting the defaults later.

The four presets you’ll use 95% of the time:

• Social short — vertical or square aspect, 5–10s, punchy motion defaults.
• Product B-roll — horizontal, 8–12s, clean motion, product-centric lighting.
• Narrative — horizontal, 10–20s, character-first, more cinematic defaults.
• Motion graphics — horizontal, 4–8s, flat colors, abstract motion.

For your first clip, pick Social short. It’s the most forgiving and the fastest to iterate on.

Write a five-part prompt

Open a new generation and type a five-part prompt. The structure:

[subject] [action] in [setting],
[lighting/mood],
[camera move/lens],
[motion detail],
[duration]

A working example for a social clip:

A barista pulls a shot of espresso at a small café counter, warm afternoon light through a window, close-up, slow push-in, subtle steam rising, 6 seconds.

Hit Generate. While it’s rendering (~90 seconds), read the next section.

Run three variations, not one

Once your first render comes back, resist the urge to judge it. Hit Generate 2 more to run two additional variations of the same prompt. This is the most important habit in AI video work.

Why: every AI video model produces meaningful variance between seeds. The best output is rarely the first one — it’s usually the second or third. If you only generate once, you’re shipping whatever you happened to get. If you generate three times, you’re picking from a sample.

While those two run, note what you like and dislike about the first render. Specifically:

• Is the subject correct? If no, your subject description needs more specificity.
• Is the action correct? If no, your action verb needs more detail (e.g., “pulls” vs “carefully pulls a full shot”).
• Is the camera move right? If no, your camera line needs adjustment.
• Is the lighting the mood you wanted? If no, rewrite the lighting clause entirely.

Pick the best, refine with image-to-video

Three renders in, pick the one closest to what you want. Even if it’s not perfect. Then click Extend with image-to-video.

This step separates hobbyists from professionals. You’re taking your favorite frame from the best generation and using it as the input for a new generation. This is how you:

• Fix a subject that drifted in the second half of a clip
• Add a specific motion the first render missed
• Extend a clip beyond the length of a single generation

A working refinement prompt:

[selected frame] — camera continues the slow push-in, the barista looks up briefly, steam continues rising, warm light stays consistent, 6 seconds.

Save your prompt to the library

If a prompt worked well, save it to your prompt library (Settings → Prompt Library). Name it by use case, not by subject: studio-rotation-10s is more useful than mug-rotation, because you’ll reuse the structure for many products.

Your prompt library will become the single highest-leverage asset you build in Deevid. Start populating it on day one.

Configure two things before you move on

Before you close the first session, set these once and stop thinking about them:

• Default aspect ratio. If you mainly post on Instagram or TikTok, set vertical as default. If you’re shipping for the web, horizontal. Changing this per-generation wastes time.
• Default render tier. Set this to match your plan. On Pro, leave it at 4K. On Starter, it’s already at 1080p.

Stop at 30 minutes

After 30 minutes, stop. Don’t try to polish the first clip into a final deliverable today. Walk away, come back the next day with fresh eyes, and iterate. The people who produce consistently strong output in Deevid are the ones who sleep on their renders.

What to learn next

Three tutorials to read in order over the next few days:

1. Anatomy of a great prompt — the deep dive on the five-part structure.
2. Character consistency workflow — once you’re comfortable with single clips, learn how to lock a character across multiple shots.
3. The 37 prompts post — a library of tested prompts to reverse-engineer.

You’ll be shipping publishable output inside of a week. That’s genuinely how fast this curve is.