You can get all cDNAs, all isoforms and then, as there should be some size differences in isoforms, make a looooong gel and cut out the right isoform from the gel after separating them properly form each other? Or, if the size differences are not big enough, You can check if there in a restriction site which is not present in Your desired splice-variant and is present in the non-desired ones and do the digestion, electrophoresis and gel isolation of the right isoform. Or, even, if there is such a restriction site, clone everything (in a plasmid which does not contain such restriction site) and then digest after ligation, so the plasmids containing non-desired isoforms are linearized (no transfection) and only the right one will stay circular - so the only clones You will get are the right ones.

Can you design specific primers? what's the difference in size?

One, easiest but most expensive, option is to get it synthesised.

You can get all cDNAs, all isoforms and then, as there should be some size differences in isoforms, make a looooong gel and cut out the right isoform from the gel after separating them properly form each other? Or, if the size differences are not big enough, You can check if there in a restriction site which is not present in Your desired splice-variant and is present in the non-desired ones and do the digestion, electrophoresis and gel isolation of the right isoform. Or, even, if there is such a restriction site, clone everything (in a plasmid which does not contain such restriction site) and then digest after ligation, so the plasmids containing non-desired isoforms are linearized (no transfection) and only the right one will stay circular - so the only clones You will get are the right ones.

Thank You very much for answering. I even designed primers in exons present only in full-length transcript and trying to combine overlapping fragments. i have to use 5 fragments to join

If you have the possibility to design a unique forward primer for your longest transcript, then it should be possible to amplify it via PCR, even if the percentage is relatively small. Optimizing such a long PCR is a different thing, that can also take a while.

thank You again for responses. I designed various variants of primers. Desirable  form is present in minority and most widespread variants are preferentially amplified. So  variants to synthesize, use of restrictions and get already cloned gene are interesting :)