XaiJu
Karl Bernard

Karl Bernard

patreon


Karl Bernard posts

Bug with Conversation Images Fixed

On Tuesday, we unfortunately introduced a bug in the Conversation Image feature. Every time an image would get generated, a new conversation would get forked. The bug was not visible at first sight as you would see your conversation history, however on reload, the entire conversation would disappear only showing the image. So the next text generation wouldn't consider your chat history.

We've fixed that issue this morning and conversations can now continue as expected after an image gen...

View Post

Conversation Images with Private Chatbot

For the Conversation Images feature to work, we need to fine-tune a specialized AI model to teach it to generate similar images than your Chatbot Image. That process takes some time (about 7 minutes per bot), therefore we've created a training queue.

All our I'm All In subscribers can now request the training of any untrained chatbot so that it can be available for Conversation Images. This includes training your private bots, or any public bot that has not been trained...

View Post

Favorite Chatbots feature now released!

We're happy to announce the release of Favorite Bots feature!!

You can now easily flag your favorite bots using the ♥️ that appear on the character card. You will then find those bots from the Favorite Chatbots menu.

Your Favorite Chatbots page is fully searchable, filterable and sortable to make it even easier to find the one you are in the mood for.

Enjoy!

View Post

Outage Resolved!

We experienced a failure from our authentication provider on Sunday from 1:45 a.m. EST to 3:05 a.m. EST. During that period of time, everyone got logged out of our service and couldn't log back in.

The issue is now resolved, we apologize for the disruption.

View Post

Linking your Patreon Account

One of the most common issue for new subscribers is linking their Patreon and SpicyChat account. That doesn't always go smoothly when emails are different, or apple Hide My Email is used or our emails don't make it to your inbox.

We've now created a self-service page that you can use if you have any issues with your pledge benefits not being reflected into your SpicyChat account.

Link your Patreon Account: https...

View Post

Conversation Images Now Available for I'm All In tier

We are happy to finally release our new conversation images feature!

How it works?

We've fine-tuned a custom AI model for each of the bots that are available for this new feature (over 50,000 public bots). This fine-tuning have allowed the AI to learn how to generate similar images, however because this is done on only 1 image, it might also attempt to replicate the style of the image or other details of the image.

When you request for an image to be generated within a con...

View Post

Conversation Images Generation Coming Soon!

While we've been quiet about feature updates, we've been working in the background on some cool things.

We expect to be releasing Conversation Images Generation in 10-14 days. While the initial results are not perfect, the feature will certainly be entertaining!

Here's in a nutshell how it works: We train a custom diffusion model on each of the chatbot using the one Avatar image that we have. When you have a conversation with that chatbot, it will use the ph...

View Post

Farewell to SpicyLlama... For Now.

Thank you for the feedback provided on the switch from the default model, ChronosHermes-13b, to our custom-tuned SpicyLlama2-13b. The feedback was mixed.

We pride ourselves on making data-driven decisions and not relying solely on first impressions. After reviewing the user submitted feedback from the past week, we've decided to revert to the previous model.

The default model has now returned to ChronosHermes-13b. We plan to reintroduce SpicyLlama in the coming we...

View Post

Meet SpicyLlama2: Our Improved AI Model

After several weeks of rigorous testing, we are proud to announce that our first custom model, trained using our proprietary data, is now becoming our primary model. This model has been shaped and improved based on the labeled data we gather each time you provide feedback on the responses.

Introducing SpicyLlama2, a refined version of Llama2-13B. We've been running this model on a random selection of generation requests to accumulate ra...

View Post

New Advanced Avatar Generation

This week we've made improvements to our avatar image generation cluster, increasing speed and reducing generation errors.

We're also introducing new advanced features:

  • Negative Prompt: better control of what you don't want in your images
  • Image Guidance: generate an image similar to another image
  • Seed Locking: Allow to redo the same generation but with changes to the prompt

Enjoy and let us know how you like it!...

View Post

Empty Generation Issue Now Resolved

After doing an update to the platform on Wednesday, 4pm EST to allow testing new AI models, we introduced a bug that resulted in about 25% of text generations to be empty.

The issue was identified and fixed at around 12pm EST today (Thursday), so during 20 hours, we generated 694,435 empty responses.   

We apologize 694,435 times for not catching and fixing that issue faster.  

Thanks for your patience and continuous support!

View Post

Longer Context Available: 4096 tokens

We're committed to providing a high-quality, free service to our user community, and it's the generous support of our premium subscribers that makes this possible for everyone.

We're excited to announce an important upgrade exclusively for our True Supporter ($14.95) and I'm All In ($24.95) subscribers. Over the past few weeks, we've integrated more advanced GPUs and leveraged recent advancements in the field. As a result, we're now able to offer a View Post

Personalized Bot Recommendations!

Hey everyone, good news! We've just rolled out our Personalized Content Recommendation feature.

What's new?

  • Get bot recommendations based on your likes and interactions.
  • Discover more of what you love, with less effort.

Our goal has always been to enhance your experience and make it as enjoyable as possible. We believe this new feature takes us one step closer to that vision.

Give it a whirl and let us know how it feels! We're al...

View Post

How to get help on SpicyChat

Submit a Ticket

You can fill the form by following the link below and we'll get back to you as soon as possible.

https://spicychat.freshdesk.com/support/tickets/new

Live Support

You can get live support from our community using Discord which is a messaging platform. Discord is also a great way to get involved with the SpicyChat.ai community.

2023-08-23 13:26:52 +0000 UTC View Post

Beta Testing Semantic Memory

We've enabled the first iteration of Semantic Memory for all paying subscribers. If you notice any issues, please report them on Discord.

How does Semantic Memory works?

The AI models that we currently use have a limit of 2048 tokens (around 1500 words) that can be used to prompt and get a reply from your chatbot. That limitation means that we can only keep a recent portion of your conversation when generating a reply.

What Semantic Memory do is identify...

View Post

Message Feedback Labeling

This week, we've added a Star Rating system that allows you to give us feedback on the quality of the text generation our models generate. This will allow us to better filter out bad data when fine-tuning future models. Thanks to those of you that will take the time to rate responses.

We've also made improvements to our text generation clusters, allowing us to now support more simultaneous users with faster generation time.

We started working ...

View Post

Message Editing Now Available!

You can now edit messages from your conversations. That a great way to fix a generation error or to steer the conversation in the direction you want.

We clarified our Community Guidelines and also added Male / Female tags this week as well as a few minor fixes.

Follow us on Twitter if you are not doing so already.

View Post

Technical Issues

Our cloud provider AWS is having technical issues which is causing some portion of the site to not respond. It's likely affecting other websites around the web.

We're hoping the issues will be resolved quickly, youc an follow the status on Amazon health status page.

https://health.aws.amazon.com/health/status

View Post

Skip the Line!

I'm happy to announce that our new Waiting Queue feature is now live!

As a regular users, you can now navigate throughout the site as you wait for your turn. Also, you can now safely reload the page without loosing your turn, as long as you stayed active in the previous 10 minutes.

As a Patreon supporter, you now completely skip the line!

Thanks for your support!

View Post

Priority Generation Queue for Supporters

Today we rolled out many updates, including Priority Generation for Patreon supporters. Message should now generate much faster when you are logged in and are a supporter.

Other improvements released today:

- Ability to delete your Characters
- You can now use your own OpenAI Key if you prefer.

We'll start working soon on a Skip the Waiting room feature.

View Post

Quick Update

I feel things are stable now, we  now have 13 servers in front of two load balancers that manage the generation queue.   

Today I fixed errors with the Generate Avatar that was often failing. Also added a limit of 30 avatar generations per day.  

Now that things are stable, I'll be taking it easy this weekend, so you won't see me around much.  

The priorities I have in mind for next week:

  • Bring your own "openai api key": That will impr...

    View Post

Message Deletion

Hi guys, thanks for all the new supporters!

Today we deployed the new Message Deletion feature, it was probably the most frequently feature!

Also happy to report that our generation backend is now much more stable, you shouldn;t be seeing as many errors, if you encounter them at all.

View Post

Midweek update!

Good afternoon! Over the last days, we've made good progress stabilizing the servers and it should soon get better with less errors once I finalize the configuration of the load balancer/reverse proxy.

We now have 12 x A100 running, at a cost of around $10k per month.

In 5 days, we grew from 0 to 20,000 visitors a day and 11,000 registered users. There's now 1200 public characters out of 6150 characters created. More than 1.7M messages have been sent during that period of time. View Post