Hector is a multi-IFU spectrograph in phase-A building for the Anglo-Australian Telescope (AAT) using fibers. Its goal is to observe 30,000 galaxies when fully complete. It is a follow up instrument of SAMI which has 13 IFUs 15' wide. The full project will have a 2 degree field corrector and at least 3 spectrographs. The IFUs are hexabundles 15' to 30' wide made of bare fibers with no buffer tightly packed and fused together to minimize the surface losses between their cores. Many different transparent spectrograph designs were studied covering a large parameter space. An important trade-off study was between the use of microlenses on the slit or just bare fibers. Microlenses have disadvantages but permit considerable simplification of the collimator by making the beam very slow. With microlenses, the collimator can be a unique spherical plano-convex lens significantly smaller than the mirror that would be needed in a reflective design. In the first part of the design, 26 different cameras where designed to cover the parameter space for 2k x 2k, 2k x 4k, or 4k x 4k detectors, and for 50, 75 or 100 micron fiber cores, with or without microlenses, with a triplet in the camera or a doublet plus singlet, and with a maximum wavelength of 1 or 1.05 microns. Not all combinations were designed but for each parameter there are at least two representative cameras with all other parameters identical. A preliminary cost estimate was made for the most promising designs which permitted to reduce them to 3 set of parameters for detailed designing, then to 2 with 4k x 4k detectors and 100 micron fiber cores. One design was made to be simple and low risk with only singlets in the collimator and cameras, the other complex with a mirror in the collimator, doublets and triplets, and innovations that could give better performance at a lower cost but with increased risk. For similar cost, the safe design has lower performances but also lower risks than the complex design. In the final stage of design, a trade-off hybrid design that keep the best of both ended being much cheaper, lower risk and with a shorter schedule than the complex design and with much better performances than the safe design. For this design, a model of the cost using lens diameter, glass cost and asphere complexity as the parameters was directly included in the Zemax Merit Function leading to significant cost improvement at fixed performances. The final design has 2 cameras covering 372 nm to 778 nm at an average resolution of 0.12 nm and able to accommodate about 1000 fibers of 1.6' core diameter. The collimator has 4 lenses and the cameras 4 and 5. There are no microlenses on the slit. The spectrograph can be upgraded with a third camera extending the coverage to 1000 nm.