The primary factor for homology modeling is the availability and the quality of template. You can try SwissModel for homology modeling, It automatically searches for the templates and you can generate a lot of models using a lot of templates and chose the best model according to the QMEAN score. SwissModel can also generates quaternary assemblies and add the prosthetic groups (such as heme) and metal ions (structural and functional) based on the template, which is a huge advantage for modeling multimeric proteins with metal ions and prosthetic groups.

You can also consider hybrid approach for proteins that does not have complete or good quality templates. I-TASSER (up to 1500 amino acids) and QUARK (for up to 200 amino acids) are good for that.

Use NCBI-BLAST and check for top entries (homologs) with PDB structures available. Use pyMOL to read these structures, provided they have with high sequence identity.

As stated above: if you have template with high similarity, it might work. The higher the similarity, the better the possibility of creating/designing an accurate model. If the similarity is low you can try I-TASSER, but you will have a VERY hard time convincing anyone that your model is reliable (particularly manuscript reviewers).

In both cases there's still need for validation, which in reality can be only done by experimental methods.

Homology modeling can be done using online servers ( e.g. SwissModel) or software (e.g. MODELLER that is available for free). In order to start generating models with either, you're going to need the following:

1) The sequence of your target protein; which could be found on UniProtKB.

2) A template with: A) a good "sequence similarity" to the target, and

B) a 3D structure submitted to the PDB. It's better to check with a professional or your supervisor on this one. Note that the structure should not have any missing residues. You can check the latter using free molecular viewers like Chimera or PyMol. A number of other factors should be considered too; like the "resolution" of your template. 

Depending on your type of research, the above-mentioned "sequence similarity" may vary. I suggest you start with a similarity of 30% or more. In order to find this template, you can use the BLAST tool on UniProt and search the database.

Now that you have a target sequence and a proper template, it's time to start building models with an online server or a software. After the process is finished, the software/server will rank the models so you can find the best one (i.e. the most accurate one) and use it.

You should know that the process is far more complicated than this. But these points can help you have a clue. Good luck! 

I agree that SwissModel is the best tool.

You have to find a good template or templates.

You can also model splitting your sequence in two or three to fit different templates and then put together your resulting models.

For visualization and analysis I recommend UCSF Chimera.

Good luck !

To perform Homology modelling you can use online servers like SwissModel, I-Tasser, Phyre2 or by using tool Modeller and UCSF Chimera . Use NCBI-Blast to find the highest identical sequence to model your query tertiary structure,

I agree with others, but If you don't have a homologous structure available then you might want to use protein threading. I would suggest using eThread.   

Visit

https://swissmodel.expasy.org/