A: How to record Esperanto voices?
Duolingo shows a perfect model of Esperanto voices.
Alas however, that was a result of a professional speaker, and it's not easy to access such a quality service, for evolving project like Memlingo.
Naturally we need some automated voice generator.
Quality Esperanto voice generation is available from these three sources.
(1) www.vocalware.com - a text-to-speech service, which also includes Esperanto voices. Cxi tio estas plej bona, gxis nun.
Malfelicxe cxi tiu parolilo havas iom da ne-ignoreblajn erarojn en kelkaj vortoj de kiuj inkluzivas 'ha', 'ho', 'hi', 'hu' -- ili estas prononcata al 'da', 'do', 'di', 'du'.
Voice quality from (1) is somewhat 'robotic', but it's quite good and clear enough as an educational material. Vocalware would be perfect if it removes following 'bugs'.
** 'ho', 'hu', 'ha' is pronounced as 'do', 'du', 'da'. This is definately a bug, but the site is not willing to correct it, even after repeated bug reportings.
** 'x'-style words are understood perfectly by this service, but if we give some UTF-8, real accented letters of Esperanto, that system produces very weird sounds, showing a lot more bugs. Definitely the system was tested only by using 'x'-system.
(2) Esperanto-sono tre similas al kelkaj europaj lingvoj, kaj iu programisto (Martin Rue) disvolvigis suficxe bonan parolilon el la pola lingvo. https://parol.martinrue.com/ Malgraux iom da pola akcento, gxi estas suficxe bona parolilo.
(3) Ekzistas malfermata publika projekto por kolektado de Esperanto vocxo. https://voice.mozilla.org/eo Kaj do ni certas, ke finfine ni havos belan Esperanto-parolilon. Cxi tiu retejo, gxis nun, nur kolektas vocxojn por trejni siajn komputilojn. Ne publikigis uzeblan parolilon. La plena datumo estas havebla sen-page!: https://voice.mozilla.org/eo/datasets
(4) translate.google.com - its Esperanto voice is useless, but some other languages with Esperantic pronunciation is also usable to produce some Esperanto voices. Estonian, Italian, Romanian voices are quite similar to Esperanto voices.
(5) another TTS voice solution is generated by using the voice from Spomenka Stimec (https://en.wikipedia.org/wiki/Spomenka_%C5%A0timec) Her voice is somewhat east-europian, but it's one of the best sources of automated Esperanto voices. (This is available on some PC software for blind people.)
So (1) and (2) are most suitable for creating learning material of Esperanto.
You can see the captured voices for the learning material from this link.