Roadmap to 0.4.0 - Githubissues

jhudsl / ari

:dancers: The Automated R Instructor

https://jhudatascience.org/ari/

Other

146 stars 37 forks source link

Roadmap to 0.4.0 #49

Open seankross opened 1 year ago

seankross commented 1 year ago

[x] get rid of text2speech specific code
[ ] specify/document voice engine API (https://github.com/jhudsl/ari/pull/57)
[x] generally reduce number of arguments in functions (especially relating to ffmpeg)
[ ] write a vignette about what kinds of args to pass to ffmpeg
[x] See if the burning subtitles functions work.
[ ] Should write some new tests for Ari that use tts coqui since it's free. A really cool test would go from tts to stt. (likely these only get run locally and not on cran)

Bonus:

[x] spin out ariExtra functions into separate utility packages

seankross commented 1 year ago

 function(images, paragraphs,
                     output = tempfile(fileext = ".mp4"),
                     voice_engine = seantalk::talk,
                     subtitles = FALSE,
                     duration = NULL,
                     engine_args = list(helium = TRUE, service = "seanbox"),
                     key_or_json_file = NULL,
                     ...)

wav <- do.call(voice_engine, args = engine_args)
# check that it's a list with a named object called wav that contains a tuneR::Wave
wav <- reduce(wav$wav, bind)
wav <- pad_wav(wav, duration = duration[i])

howardbaek commented 1 year ago

@cansavvy This is a roadmap of the next release of ari that came out of my meeting with Sean.

howardbaek commented 1 year ago

I documented the relationship between the main ari functions here

cansavvy commented 1 year ago

I'm super excited about this. But I think since I wasn't in the meeting these bullet points lead me to more questions than me understanding the roadmap. Would love to meet with everyone involved to chat and divide up tasks/roles for this project.