lucab85 / avg Goto Github PK

It looks like the text alignment gets a little messed up on some slides so it'd be good to explore this further.

In addition to being able to specify different voices, perhaps we could allow different languages to be specified as well. I suggest this because I saw that Amazon Polly supports both voices and languages, which are technically different things.

That way, it could be possible to produce multiple different videos of the same presentation, in different languages but using an accent that is suited to each language.

Provide some additional detailed in the README about how the text-to-speech conversion is being done, so people can better understand the other values they could provide in the default.env file, for example.

In fact, perhaps we could add comments into default.env, pointing people to the relevant AWS page to see the available values for each variable.

How about creating an empty "share" folder in the AVG repo so that the step to create this (during the setup phase) is not actually needed?

For sure, people can select/use a different folder later on but for most use cases (and in the interest in increased adoption), maybe this would help?

If your slide deck used custom fonts (e.g. Red Hat Display), which are not available inside the docker container, then the resulting video may contain incorrectly formatted slides. It may be worth exploring if there's any way to add the .ttf files for the fonts in question to the ARI ecosystem.

Add some details to the README to help people understand how long it takes to render each video. Clearly, it depends on the length of the audio files produced from the speaker notes, but is it exactly 100% of their combined length or somewhere in between?

I recently switched to a new MacBook Pro powered by the Apple M1 Silicon chipset and I ran into some problems while trying to set up the AVG environment there. Specifically, I got this warning when I ran the docker run command:

$ docker run --name="avg" -dit --mount type=bind,source=$(pwd)/share,target=/share --env-file=default.env  lucab85/avg
WARNING: The requested image's platform (linux/amd64) does not match the detected host platform (linux/arm64/v8) and no specific platform was requested

Could it be that we need to build an alternative version of the the base image, that suits the M1 chipset?

A few people have commented on recent demos (that I've shared) saying that the voice we've used speaks too quickly. Personally, the voice in question (Joanna) is probably the clearest one I've heard but I'll admit that it could indeed be a little slower when talking.

It would be good if Amazon Polly allowed you to control the pitch (or speed) of the voice.