My clients love the bots I’ve made thus far and came up with a great idea!
A bot that can create a video from text or voice, and then format it to the specific social media format / px ratio.
I think this is BEAST and I could sell a lot of it right now
@chadvmm you will need to build an action to an external tool
Ok ty. Any suggestion(s) as to which one?
@chadvmm Pika for random videos or HeyGen for avatars. You will need to pay for their APIs.
Clutch thank you. When you say “random,” I’m wondering if it would ignore the content and context I provide it? I’m relatively familiar with HeyGen, I think they’ve definitely improved. thank you so much once again, loving it here.
What do you think about this AI? Video Generation APIs?
@admin_mike , have you heard of Tavus?
If you want **Video Avatars, you could build an Action for it using the APIs you mentioned. It will render a video in the chat that users can hit play.
If you want a consistent video avatar that end-users just talk to, that is a fundamental change in modality (going from text-based conversation to a video/audio conversation). We’ll eventually build that, and most likely will do so outside the context of Actions. It will need to be a separate builder (like Forms, Chatbots, Video Avatars). Pretty big feature we won’t be tackling in the short-term, but will be doing in the long-term.
I love the idea of it rendering and being seen right there and then.
What would you say the best API would be to create videos that aren’t necessarily avatar specific
In my particular scenario that would be related to rehab, strength training and performance
If you want text-to-video generation, the area is very limited. There are tons of models that can generate videos. But if you want the video to show something very specific, you will struggle to get good results.
If you want to text-to-video AI models, can check out:
If you want avatars, you can check out:
Ty so much for all of this!
Have you heard anything about this one?
@chadvmm that’s a good idea to keep everything in one place but be aware they are using open source models (LLM, Dreambooth And Stable Diffusion Models - Stable Diffusion And Dreambooth API - Generate and Finetune Dreambooth Stable Diffusion using API).
So I would rather talk to the model directly than adding an intermediary step if it makes sense.
As I think more about this process for my clients, I think ideally your Pick Axe bot would help auto generate the video right there and then for them to see prior to submitting the action. Just wanted to confirm, can this be done with the text to video AI models you’ve suggested?
Yes, the videos will be very bad though. I recommend playing around with the video models, testing out the text inputs you expect to be getting. See if you even like the videos before you configure an action.