Following Ulrike's suggestion, and my code at create latex symbol from vector graphics. To get phrase1.jpg, I just cropped the title to this page. To get phrase2.jpg, I cropped phrase1, so as to demonstrate the method when the descenders are different for each phrase (only change is the \raisebox argument).
The scalerel approach adopted will scale the graphics to the height of a letter X, regardless of the current font size.
\documentclass{article}
\fboxsep=-\fboxrule
\usepackage{scalerel}
\def\phraseA{\scalerel*{%
\setbox0=\hbox{\raisebox{-9pt}{\includegraphics{phrase1}}}\dp0=0pt\box0}{X}}
\def\phraseB{\scalerel*{%
\setbox0=\hbox{\raisebox{-2pt}{\includegraphics{phrase2}}}\dp0=0pt\box0}{X}}
\begin{document}
Can I insert my \phraseA{} inline along with \phraseB?
\tiny Can I insert my \phraseA{} inline along with \phraseB?
The original images:\fboxsep=1pt\relax
\fbox{\includegraphics{phrase1}} and \fbox{\includegraphics{phrase2}}
\end{document}

If, instead of cropping jpgs with different top/bottom cut lines, you use Xelatex to to create a standalone PDF of \strut<Tamil word or phrase> with minimal margin, then each implementation of of \phraseX will use the identical \raisebox parameters. Furthermore, the image, deriving from a vector graphic PDF, should scale without pixelation.