Practical tutorial
This walkthrough explains how to use ai3dgen from the first idea to a usable 3D result. Follow the sequence, compare each revision, and keep your prompt focused while you learn the workflow.
A small amount of preparation makes each generation easier to judge and revise.
6 min readUse the same loop each time: define the object, generate a first pass, then inspect what needs to change.
Name the main object first, then describe the shape that should be visible from a distance. Add the intended style, pose, proportions, and a few defining features. For example, “compact sci-fi rover with six wheels and a raised sensor mast” is easier to interpret than a long list of unrelated adjectives.
PROMPT RULE
Put the object, form, and defining feature before atmosphere or decoration.
If you are working from an image, decide what the reference should contribute. It may establish the front view, the overall shape, a material cue, or a distinctive detail. State that purpose in your description so the reference supports the idea instead of competing with it.
REFERENCE CHECK
One strong visual cue is usually easier to preserve than several conflicting references.
Treat the first output as a diagnostic pass. Check whether the silhouette reads correctly, whether the important features survived, and whether the proportions match your intention. Keep what works, then ask for one or two targeted changes rather than rewriting every detail.
REVISION RULE
Change the weakest visible feature first and leave successful parts unchanged.
This three-part loop keeps the process simple enough to repeat while leaving room for creative iteration.
Write the subject, shape, style, and two or three details that matter most.
Submit the idea and use the first result to reveal what the description communicated clearly.
Make a focused revision, compare it with the previous pass, and keep the stronger direction.
Most disappointing results come from unclear priorities, overloaded descriptions, or revisions that change too much at once.
Name one primary subject before listing secondary objects or scene details.
RequiredThis gives the generation a clear visual priority.
Describe visible form, pose, and proportions instead of relying only on mood words.
RequiredConcrete shape language is easier to assess in the result.
Keep the first prompt short enough to identify the most important requirements.
RequiredAdd complexity after the main silhouette works.
Use a reference image when a specific structure, angle, or material is difficult to explain.
OptionalA reference is helpful, but it should have a defined purpose.
Review one change at a time when a revision misses the mark.
RequiredSmall revisions make the cause of improvement easier to understand.
Once the basic workflow feels familiar, these habits can make your iterations more consistent and easier to compare.
Use the divider to compare a rough first pass with a more focused revision, and look for changes in silhouette, detail, and balance.
First pass
Refined pass
Compare the same subject before changing direction.
Start with one object, one clear visual goal, and a small number of defining details. A focused first attempt gives you a useful baseline for the next revision.
Start creatingThese answers cover the practical questions that usually come up during a first session.
Start with the main subject and its visible silhouette. Then add the style, pose or proportions, and only the details that are essential to recognizing the object.
No image reference is required for a text-led starting point, but a reference can help communicate a specific shape, viewpoint, material, or distinctive feature. Use it with a clear purpose rather than adding several unrelated images.
The prompt may contain too many competing priorities or rely on abstract language. Simplify it, put the subject and most important form first, and revise the weakest visible feature instead of changing everything.
Begin with the subject, overall form, style, and two or three defining details. Once the silhouette and identity work, add secondary features in later iterations so you can tell which change affected the result.
Compare the result with your original goal and identify one concrete shortfall, such as proportions, pose, or a missing feature. Make a targeted revision, then keep the version that better preserves the subject and the successful parts of the earlier pass.