> ## Documentation Index
> Fetch the complete documentation index at: https://docs.holala.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Common Mistakes

> The input mistakes that quietly ruin holala results, and how to fix each one before you generate.

Most disappointing results trace back to one thing: the input photo, not the AI. holala refines what your photo already shows — it does not invent angles, details, or context that aren't there. Fix the input, and the output follows.

Here are the mistakes that come up most often, and the fix for each.

<AccordionGroup>
  <Accordion title="Wrong input angle">
    You uploaded a three-quarter shot but expected a straight-on, front-facing result. holala can restyle a scene, but it cannot rotate a product to an angle your photo never captured.

    **Fix:** Match your input angle to your desired output angle. If you need a front-facing marketing image, shoot the product front-facing. See <Icon icon="camera" /> [Input & Uploads](/input-uploads) for angle guidance per template.
  </Accordion>

  <Accordion title="Busy or patterned backgrounds">
    A cluttered backdrop, a patterned surface, or a busy room behind the product competes with the AI's read of the product itself, and often bleeds into the generated scene in odd ways.

    **Fix:** Shoot on a plain, neutral background whenever possible. The simpler your backdrop, the cleaner holala's read of your product edges, color, and shape.
  </Accordion>

  <Accordion title="Low-resolution or cropped product">
    If the source image is small, blurry, or cropped so part of the product is cut off, holala has less real detail to work with and has to guess at what's missing.

    **Fix:** Upload the highest-resolution photo you have, fully framed with the whole product visible. Check <Icon icon="upload" /> [Input & Uploads](/input-uploads) for minimum recommendations before you generate.
  </Accordion>

  <Accordion title="Multiple products in one frame">
    Uploading a photo with two or more products together confuses which item the template should feature, and Templates are built as one-job functions for a single product.

    **Fix:** Upload one product per generation. Run separate jobs for separate products, then organize the results in your Library.
  </Accordion>

  <Accordion title="Using 1K for final output">
    1K is a quick test resolution. Publishing a 1K result as your final marketing image gives you softer detail than the platform is capable of.

    **Fix:** Test your template and inputs at 1K to confirm the composition works, then re-run at 2K or 4K for anything you'll actually publish. See <Icon icon="coins" /> [Credits](/credits) for how resolution affects credit cost.
  </Accordion>

  <Accordion title="A reference image that fights the product">
    An optional reference image is meant to guide style, scene, or lighting. A reference from the wrong product category, a busy or cluttered reference, or a low-resolution reference pulls the generation in a conflicting direction instead of helping it.

    **Fix:** Choose a reference that matches your product's category and is itself a clean, high-resolution image. If a reference is fighting your results, remove it and re-test with the product image alone.
  </Accordion>

  <Accordion title="Expecting the AI to invent detail">
    holala refines what's already in your photo — it does not fabricate a side, texture, or feature your input never showed. If your input doesn't show it, the output won't either.

    **Fix:** Before you generate, ask: does this input already contain everything I want the final image to show? If not, reshoot the input rather than expecting the generation to fill the gap.
  </Accordion>
</AccordionGroup>

<Warning>
  Every one of these mistakes traces back to the input photo. When a result looks off, check your source image first before assuming the template or settings are wrong.
</Warning>

## Quick Do / Don't reference

<Columns cols={2}>
  <Column>
    **Do**

    * Match input angle to desired output angle
    * Use a plain, neutral background
    * Upload full-resolution, fully-framed photos
    * One product per generation
    * Test at 1K, finalize at 2K or 4K
    * Pick a reference in the same category, clean and high-res
  </Column>

  <Column>
    **Don't**

    * Expect a front result from a side-angle input
    * Shoot on busy or patterned backdrops
    * Upload cropped or blurry source images
    * Combine multiple products in one frame
    * Publish 1K as your final image
    * Use a low-quality or mismatched reference image
  </Column>
</Columns>

## Related reading

<CardGroup cols={2}>
  <Card title="Input & Uploads" icon="upload" href="/input-uploads">
    The full guide to preparing product and reference images before you generate.
  </Card>

  <Card title="Credits" icon="coins" href="/credits">
    How resolution and variation count affect what a generation costs.
  </Card>
</CardGroup>