I currently pay $99/mo for around 10 hours of generated audio with ElevenLabs. If you're looking for a free text-to-speech alternative, you might want to check out TTS. This video explains more: ruclips.net/video/XxcuG8Yljwc/видео.html
The response time with MemGPT can vary significantly based on the LLM you use. For a free alternative to ElevenLabs, check out our latest video:: ruclips.net/video/XxcuG8Yljwc/видео.html
Excellent video, you are amazing! I have a question. In AutoGen, there are the TeachableAgents, which you showed in a previous video. I want to create an assistant like Alexa or Google, but using an LLM behind it, to give it a touch like JARVIS by adding tools such as: - control of home assistant - internet search - managing my schedule - alarms - among other things My question is, what would you recommend more? Using MemGPT or the TeachableAgents? Best regards.
Thank you for the compliment! If you want a smart assistant with capabilities like home control, internet searches, and managing schedules, I'd recommend opting for MemGPT. It's great because of its flexible input variants (e.g., via API calls). While AutoGen can stream answers using WebSockets (microsoft.github.io/autogen/docs/notebooks/agentchat_websockets/), it tends to be more complex. MemGPT offers the flexibility you need for such an advanced project
Unlimited Memory????But I'm aware of that's not possible every single AI language model I've ever interacted with seems to have a very limited long-term memory majority of them memory dump after about a week or three days. The ones on platforms like replica as an example have some memory but not much. There is even been a lot of discussions in RUclips videos about as far as I know there's nobody out there that the AI has any real permanent memory to speak of to any real extent there's always a pretty small token limit on the permanent memory
Can you use a local model?
This.
@@grymvision3094 it uses chat gpt 3.5 on their server, not a local machine so you need an API key to use it.
thank you a lot. any plan to take it to next level and explore creating a talking avatar on top of this, with open source?
What do these companies charge for API keys?
Awesome project. Will have to build this for the wife :D
Glad you liked it! Your wife will love it! If you have any questions or need any help building it, feel free to ask. 😊
nice video. thx a lot. but wenn ich close the server. all my data are gone right? all the learnings?
Isn't ElevenLabs very expensive?
I currently pay $99/mo for around 10 hours of generated audio with ElevenLabs. If you're looking for a free text-to-speech alternative, you might want to check out TTS. This video explains more: ruclips.net/video/XxcuG8Yljwc/видео.html
Memgpt is slow as it consists of agent . Also elevenlabs is free 😢 please give a free alternative ❤😅
The response time with MemGPT can vary significantly based on the LLM you use. For a free alternative to ElevenLabs, check out our latest video:: ruclips.net/video/XxcuG8Yljwc/видео.html
I'm using Groq for the speed
@@ai-for-devs I'm using Groq for the speed.
Excellent video, you are amazing!
I have a question. In AutoGen, there are the TeachableAgents, which you showed in a previous video. I want to create an assistant like Alexa or Google, but using an LLM behind it, to give it a touch like JARVIS by adding tools such as:
- control of home assistant
- internet search
- managing my schedule
- alarms
- among other things
My question is, what would you recommend more? Using MemGPT or the TeachableAgents?
Best regards.
I have made this exact but with many more features 😅 . If u r interested u can reply this 😅❤
Thank you for the compliment! If you want a smart assistant with capabilities like home control, internet searches, and managing schedules, I'd recommend opting for MemGPT. It's great because of its flexible input variants (e.g., via API calls). While AutoGen can stream answers using WebSockets (microsoft.github.io/autogen/docs/notebooks/agentchat_websockets/), it tends to be more complex. MemGPT offers the flexibility you need for such an advanced project
@@lokeshart3340 I am most intrigued!
@@lokeshart3340 Yes, I'm interested!
@@lokeshart3340 yes, I’m interested
Unlimited Memory????But I'm aware of that's not possible every single AI language model I've ever interacted with seems to have a very limited long-term memory majority of them memory dump after about a week or three days. The ones on platforms like replica as an example have some memory but not much. There is even been a lot of discussions in RUclips videos about as far as I know there's nobody out there that the AI has any real permanent memory to speak of to any real extent there's always a pretty small token limit on the permanent memory
Please watch the video ;-) (memgpt uses a permanent vector store)