Feature Requests

all my jobs are queued forever! nothing being processed! get_job_predictions()

As a user's number of projects grows, the need to have some means to sort through them to find specific projects will increase. I'm probably one of those users who will need a means to sort through my projects soonish. There's all kinds of ways to supply sorting through projects. Keyword search the titles, created date, modified date, tags, etc. In my use case, a keyword search of project titles would help. I know what project I'm aiming for, so just being able to type into a field and have the project list filter down would work for me. My use case: I am working on audiobooks. I have one fantasy series across multiple books. I cannot have one project per book because my chapters are too long. I use one "project" for each of my books' chapters (10-20k words), and one "chapter" per scene (800-2500 words). Therefore, I have a bunch of projects which I name "Book - Chapter XX" My project list looks something like this right now: * Lorem Ipsum - Ch 2 The Fate of Sunshine * Lorem Ipsum - Ch 1 Who Would Have Thought * Lorem the Bar - Ch 1 * Lorem a Foo - Ch 2 * Lorem Escobar - Ch 5 * Lorem Escobar - Ch 1 * Lorem the Bar - Ch 5 * Lorem a Foo - Ch 1 As you can see, ordering my projects by "created date" just makes for a random list for me. If I had a field that would filter the list, I could type in "Escobar" and it would show me all of my "Lorem Escobar" chapters.

I would fully switch to using EVI3 if it had a knowledge base in RAG. I'm currently building a company and spending thousands of dollars on AI, and I think this is the best.

This is a big ask, but also a big help for audiobooks. especially of a fantasy setting. It'd be great to be able to have accent relationships between an existing voice(s) and the new voice being generated. Like if during the voice creation process I could reference an existing voice and say that the accent should be similar. This would make it more straightforward to design voices around families and regions. For example: I could say that two voices are siblings, so their accents would be identical. Or, I could say that the two voices are from a similar region, so their accents would be related. It'd also be handy to say that two voices are NOT related, and therefore help to make them distinct from each other.

Would it be possible to upload a file with the texts/audios of several participants and obtain the individual analysis for each participant?

In the Christmas Lighting hobby, we stand up Mega trees that talk to the crowd and if you have more than one tree, talk to one another. To make it more real, allowing the two AI voices to overlap in a conversation would make the mega trees more real and life-like.

It would be nice if I could manually set the pausing between blocks, but manual pauses are stripped from the beginning and ending of blocks making that impossible. There are times when an extra breath is needed between blocks, just a little more than normal. Usually when the topic is changing just enough where a breath can help orient the listener to knowing that the thought has concluded and now a new thought is being started. But with the stripping of pauses from the beginning and ending of blocks, I can't manually add this breath in the app. I end up being required to post-process the export in order to add in these pauses.

One approach is to have some UI on that first screen where you can choose "current chapter" or "select a specific chapter". Another approach is that when choosing to download a single chapter, the current active chapter is automatically selected. I export each chapter as I finish it. As the quantity of chapters grows, it gets increasingly annoying to have to page through the list of chapters just to be able to export the current active chapter.

It would be amazing if you could add a RAG support directly on the platform for adding specific knowledge for the EVI. Thanks!

Under projects -> Settings there is a list of voices used in the chapter. It would be nice if the current utterance selection's selected voice was indicated/highlighted in the "select used voice" list. I frequently have the "Acting Instructions" box open which has the sidebar scroll so I can't see the top of it, hiding where the current active voice is displayed. I know you have colors to indicate which voice is currently active, but they don't work well for me with my red screen tinting (f.lux / blue light filtering). I imagine similar color differentiation issues may occur with people who are color blind.