Weekly summary, Jan 11-17 2024

This week, I have been concentrating on two items: better control of TTS, and an initial version of “Simple C-LARA”. Here are the details:

TTS

In response to Cathy’s post, pointing out that our default Google TTS isn’t very good, I did some work to integrate the OpenAI TTS engine, allowing the user to select the voice they want. The details are here. This is indeed a substantial improvement for English, and possibly for French, but I’m not so impressed with performance in other languages; as OpenAI say, it’s been optimised for English. Cathy looked around some more, and it seems like we should be thinking of getting a license for ElevenLabs. Let’s discuss.

“Simple C-LARA”

We’ve been kicking around for some time the idea of “Simple C-LARA”, a slimmed-down version of the interface where most of the options are hidden and you can produce a piece of content with minimal effort. I have a preliminary implementation working end-to-end on my laptop and should be able to install it on the UniSA server by Friday so that people can experiment. The content creation process is as follows:

Fill in the title, text language and annotation language, and hit “Create project”.
Write a prompt to say what text you want to generate, and hit “Create text and image”. Simple C-LARA uses GPT-4 to create the text as usual, then uses DALL-E-3 to create an image based on it.
If you don’t like the text or image, hit “Rewrite text” or “Regenerate image” to get new variants.
When you’re happy, hit “Generate multimodal text” to add the other annotations and put everything together, When it’s finished, Simple C-LARA shows you a link.
If you like the result, hit “Post to social network” to make it generally available.

It’s still very rough, but I think we will be able to improve it quickly in response to feedback. I will demo it in tomorrow’s Zoom!

Next Zoom call

Thu Jan 18, 2023, 20:00 Adelaide (= 09.30 Iceland = 09.30 Ireland/Faroe Islands = 10.30 Europe = 11.30 Israel = 13.00 Iran = 17.30 China = 20:30 Melbourne/New Caledonia)