Two of the single most common reasons a text extraction comes back inaccurate are an angled photo and poor lighting — both of which are entirely within your control to fix before you even upload the image, rather than something the tool itself can fully compensate for afterward.
Why angle distorts text recognition
OCR systems are built around recognising letter shapes as they'd normally appear — upright and proportioned correctly. A photo taken at a meaningful angle distorts those shapes, stretching and skewing letters in a way that makes them harder to match against the expected patterns the recognition system relies on.
Why low light causes its own separate problem
Poor lighting reduces the contrast between the text and its background, which is exactly the distinction OCR depends on to identify where letters actually are in the first place. Dim, uneven, or shadow-heavy lighting makes this boundary less clear, directly reducing how reliably the text can be identified at all, separate from any issue with the text's shape or angle.
Steps to avoid both problems before you start
- Position yourself directly in front of the text, as straight-on as practically possible, rather than at a noticeable angle.
- Find or create even, reasonably bright lighting, avoiding both deep shadow and harsh direct glare.
- Hold the camera steady to avoid motion blur, which compounds both of the issues above.
- Check the photo looks clear and readable to your own eye before uploading — if you're struggling to read it yourself, OCR will likely struggle too.
What to do if you can't avoid an angle or poor lighting
If the source material genuinely can't be repositioned or better lit — a sign fixed to a wall, for instance — take several photos from slightly different angles and lighting conditions if possible, and try extracting from each, since one might come through meaningfully more accurately than another even for the same fundamentally difficult source.
Why a quick re-shoot is often faster than fighting a bad result
If an initial extraction comes back poor and you suspect angle or lighting is the cause, it's often faster to simply retake the photo more carefully than to try to manually correct a heavily garbled result, particularly for a longer passage of text.
Combining both issues makes accuracy drop further
A photo that's both angled and poorly lit compounds these two separate problems, often producing a meaningfully worse result than either issue alone would cause. Addressing both — angle and lighting — together, rather than fixing just one, gives you the best realistic chance at an accurate extraction.
Using your phone's own camera features to help beforehand
Many phone cameras include a built-in document or flat-surface detection mode that helps automatically correct minor angle issues and improve exposure before you even take the photo, which is worth enabling if your phone offers it and you photograph text regularly.
Why a slightly blurry but well-lit, straight photo beats a sharp but angled one
If you have to choose between two imperfect options, a straight-on, well-lit photo with slight blur generally produces a more accurate extraction than a perfectly sharp photo taken at a significant angle, since angle distortion tends to cause more fundamental recognition problems than mild blur.
Why this knowledge helps even outside of OCR specifically
Understanding what makes text hard for a machine to read also tends to make it genuinely harder for a human to read quickly too; applying these same principles — straight angle, good lighting — benefits general photo-taking of any text-containing subject, not just situations where you plan to run OCR.
Building this into how you generally photograph documents
Once these principles become habitual, they tend to improve the general quality of any document photo you take, not just ones specifically intended for OCR, since straight, well-lit photos are simply more useful and legible documents overall.
Practising on a few test photos if accuracy really matters to you
If you're preparing to extract text from something important, running a couple of test photos through the tool first, deliberately varying angle and lighting slightly, helps you calibrate what level of care your specific camera and setup actually require for a reliable result.
Frequently asked questions
Why does photographing text at an angle reduce OCR accuracy?
It distorts letter shapes away from how the recognition system expects to see them, making accurate matching harder.
Why does poor lighting cause problems separate from angle?
Low contrast between text and background makes it harder for OCR to even identify where the letters are in the first place.
What's the simplest way to improve my chances before uploading?
Photograph straight-on, in even, reasonably bright lighting, holding the camera steady.
What if I genuinely can't reposition the source material?
Try a few photos from slightly different angles and lighting conditions, since one may come through more accurately than another.
Is it faster to fix a bad photo or retake it?
Retaking it is often faster than trying to manually correct a heavily garbled extraction result.
Do angle and lighting problems compound each other?
Yes, a photo with both issues together often produces a meaningfully worse result than either issue alone.
Is Image to Text free to use?
Yes, no sign-up required.