Skip to main content
Logo image
Apptastic Coder
AI Tools

PixelDojo Tutorial: Learn AI Image and Video Generation from Scratch

Create a product image, refine your prompt, and turn the result into a short video

By Team Apptastic8 mins read

Full Article

Read the complete article, including examples, background, practical insights, and technical details.

PixelDojo image and video generation tutorial with purple abstract prism artwork

Affiliate disclosure: This tutorial contains affiliate links. We may earn a commission if you purchase through them.

Labels: PixelDojo, AI Image Generation, AI Video Generation, Beginner Tutorial, Prompt Engineering.

Creating your first AI image is exciting. Getting an image you actually intended, then turning it into convincing motion, takes a more deliberate process. You need a clear brief, a useful prompt, and a way to evaluate what the model produces.

This tutorial teaches that process through one small project: a promotional image of a fictional ceramic coffee cup, followed by a short animated shot. You can adapt the exercise for a portfolio, classroom project, product concept, or social post.

The workflow below is based on PixelDojo's public documentation, checked on November 22, 2026. The prompts are original practice exercises, not claims of tested output. Controls and available models can change.

Explore PixelDojo — affiliate link.

1. Understand the three generation workflows

PixelDojo brings image generation, video generation, editing, and enhancement tools into one workspace. Its official overview describes workflows that start from either text or an existing image.

Text to image creates a still picture from your description. Think of it as giving a photographer or illustrator a brief: subject, surroundings, composition, lighting, and visual style.

Image to image uses an existing picture as a starting point. Depending on the selected tool, you can request changes to its appearance, surroundings, or details.

Image to video adds motion to a starting image. Your instructions describe what moves and how the camera behaves. Text to video instead asks the model to invent the scene and its motion together.

For this exercise, begin with a still image. Separating composition from motion makes mistakes easier to diagnose.

Later, try text to video with the same scene. Combine your image description with a simple motion instruction, then compare the result with your animated reference image. Notice which approach gives you better control over framing, object appearance, and movement. This comparison teaches when to establish a visual starting point and when to let the model invent the whole shot.

2. Set up a manageable first project

Open PixelDojo, create or access your account, and inspect the available tools and generation costs before starting. Choose an image generator from the image tools. You do not need an API key, programming environment, or custom model training for this exercise.

Use this brief:

  • Subject: one matte teal ceramic coffee cup.
  • Setting: a wooden table beside a softly lit window.
  • Mood: calm, warm, and realistic.
  • Deliverables: one landscape image and one short animated clip.
  • Constraints: no people, lettering, logos, or additional cups.

Choose a landscape aspect ratio, such as 16:9, if the selected tool supports it. For a vertical social project, choose 9:16 from the beginning instead. Changing formats later may crop away useful details.

Start with the default settings and one output. Learn what each control changes before increasing batch sizes or resolution.

3. Write your first image prompt

A useful prompt follows this structure: subject, setting, composition, lighting, style, and constraints. Specific visible details usually help more than a long string of adjectives.

Try this original prompt:

A single matte teal ceramic coffee cup on a light oak table beside a window. Three-quarter view at table height, cup positioned slightly left of center, empty space on the right. Soft morning light from the left, gentle shadows, warm neutral background, realistic product photography, shallow depth of field. Plain unbranded cup, no lettering, no people, no extra objects.

Select an available general image model and generate one result. PixelDojo's image generation page lists its current options; use the model description and displayed cost to guide your choice.

Keep the prompt unchanged for your first comparison between models. Otherwise, you will not know whether a better result came from the model or your rewritten instructions.

If a separate negative prompt field exists, use it for unwanted elements. Do not assume every model provides one.

4. Evaluate before you regenerate

Treat the first output as a draft. Inspect the cup's rim, handle, base, shadow, and relationship to the table. Check whether the lighting matches your description and whether the composition leaves room for your intended layout.

Ask three questions: Is the subject correct? Is the geometry believable? Does the image serve the brief?

Suppose the scene looks attractive but includes two cups. Strengthen the subject constraint: “Exactly one cup, isolated on an otherwise empty table.” If the background is distracting, request a plain wall and fewer decorative details.

Change one meaningful instruction per iteration. Keep a short record of the prompt, model, settings, and outcome. This turns random experimentation into something you can repeat.

If your tool exposes a seed, recording it can help comparisons, but a seed does not guarantee identical results across different models or versions.

5. Refine the image without restarting

When the composition works but one detail does not, explore PixelDojo's editing tools. Its image to image workflow provides a starting point for transforming an existing picture.

Upload or select your saved image in the relevant editor. Describe the intended change precisely:

Keep the cup, camera angle, and table arrangement unchanged. Make the background a clean warm beige wall. Preserve the soft window lighting and realistic ceramic texture.

Compare the edited output against the original. Even an instruction to preserve something may not prevent unintended changes, so inspect the handle and silhouette again.

Where a masking or inpainting tool is available, use it to target a small problem area. Save your favorite composition before experimenting further.

Upscale only after the image is worth keeping. Higher resolution cannot reliably repair a fundamentally malformed object; it may simply make the mistake more visible.

6. Turn your image into video

Open an image-to-video tool and upload the selected image, or use an available transfer action from your library. PixelDojo's image-to-video guide describes uploading an image, choosing an engine, and specifying motion.

Select a short duration and a modest resolution supported by that model. Confirm the displayed cost before running it. Available durations, aspect ratios, and audio options depend on the engine.

Use a motion prompt rather than repeating the entire image description:

The camera slowly pushes toward the coffee cup. A thin ribbon of steam rises gently from the coffee. The cup and table remain still. Window light stays consistent. Smooth continuous movement, subtle realistic motion, no scene cuts.

This asks for one camera movement and one environmental action. Avoid combining a rotating cup, flying beans, changing backgrounds, dramatic zooms, and multiple cuts in your first attempt.

Watch the full result before judging it. A convincing opening frame can hide a handle that bends halfway through the clip.

7. Learn to diagnose motion problems

Video introduces temporal consistency: objects should retain their identity and shape as time passes. A good still image does not guarantee a stable animation.

If the cup changes shape, reduce the requested movement and try a locked camera. If the steam becomes a large cloud, describe it as faint and slow. If the camera swings unpredictably, specify a stationary viewpoint or a gentle forward move without rotation.

Make one adjustment, regenerate, and compare. Keep the same starting image while testing motion prompts so you can isolate the change.

For a useful experiment, create two versions: one with a locked camera and one with a slow push-in. Decide which better communicates the calm mood of your brief.

Check audio separately when supported. Do not assume every generated clip includes sound, or that silence means the generation failed.

8. Finish and export your project

Download the chosen image and clip from the available result or library controls. Check the actual file dimensions and playback before sharing. Use the format offered by your selected tool rather than assuming every engine exports identical files.

Add final typography, branding, captions, or music in an editor when needed. Keeping exact promotional text outside the generation step makes spelling and layout easier to control.

For a longer sequence, create separate shots with consistent reference images and lighting. PixelDojo also describes Film Studio for assembling multi-shot projects, but a single successful clip is enough for this first lesson.

Keep your original image, final clip, and prompt notes together. Name files clearly, such as coffee-cup-image-v1 and coffee-cup-slow-push-v2, so you can identify successful experiments later.

9. Keep learning costs predictable

Check PixelDojo's current pricing and the estimate shown before each generation. Model choice, resolution, and video duration can affect credit use. Free effects and paid generation access are different offers; verify which workflow your account includes.

Set an experiment budget before you begin. For example, allow three image attempts and two video attempts, then review what you learned before spending more.

Use inexpensive drafts to test composition. Reserve larger outputs for a prompt and image you already like. Avoid changing several settings simultaneously, because an unclear experiment wastes both credits and learning time.

If your selected model offers an unlimited or relaxed mode, read its current limits instead of assuming it covers every tool or output size.

10. Build your own repeatable practice

Repeat the exercise with a different subject: a desk lamp, a houseplant, or an imaginary perfume bottle. Keep the same sequence: brief, image, critique, edit, motion, review, export.

Use your own reference material or material you have permission to upload. For real products, verify that generated details match what you actually sell. Review the applicable platform and model terms before commercial use.

Your goal is to become better at directing and evaluating output. A repeatable process matters more than discovering one impressive prompt.

Start your first PixelDojo image and video project — affiliate link.

Nov 22, 2026

Comments load as you scroll…