NEW NSFW Pygmalion 13B! THE Ultimate UNCENSORED Roleplay KING!
TLDRThe video script discusses the release of the new Pygmalion 13 billion parameter model, an upgrade from the previous 7 billion model. The creator, a fan of Pygmalion for conversation and role-play, shares a hands-on comparison using the perplexity score benchmark, revealing the new model's superior performance. Despite this, the differences in conversation quality are subjective and challenging to demonstrate. The creator engages with a character named Sandra using both models, showcasing their capabilities and inviting viewers to decide which model they prefer. The video emphasizes the fun and immersive nature of these AI conversational models, particularly when combined with the Whisper and 11 Labs extensions.
Takeaways
- 🚀 The new Pygmalion model has been released, an upgrade from the previous Bigvanion 7B model.
- 🗣️ The narrator initially didn't plan to make a video about the new language model but decided to due to their fondness for Bigvanion.
- 📊 The Pygmalion 13B model performed better than several other 13B parameter models in perplexity score benchmarks.
- 🤖 The Big Bang models are specialized for conversation and role-play, and are not meant for general question-answering tasks.
- 🎭 The narrator plans to demonstrate the differences between the two models by role-playing with characters created using each model.
- 💻 The video will use the Uber Bugatti generation web UI with the Whisper extension from 11 Labs for interaction.
- 🌐 Community-shared characters can be downloaded from the narrator's Discord server or other platforms like guru.plus.
- 📖 The role-play scenario involves a character named Sandra, an environmental scientist, and her interactions with her partner.
- 🍽️ The conversation between Sandra and her partner revolves around their day, work, and plans for the evening, including dinner and watching a TV show.
- 🎥 The narrator concludes that both models perform similarly well, with the 7B model having slightly more action cues for realism.
- 📝 The narrator encourages viewers to try the models themselves and share their preferences in the comments.
Q & A
What is the new Pygmalion model mentioned in the script?
-The new Pygmalion model is an upgraded language model for conversation and role play, specifically the 13 billion parameter version, which has just been released.
How does the new Pygmalion 13B model compare to the previous 7B model?
-The new Pygmalion 13B model performed better than the previous 7B model in perplexity score benchmarks, outperforming other models like Stable Diffusion, Kunia, and the original Vikunya.
Why is it difficult to show the differences between the two Pygmalion models on camera?
-It is difficult because conversational models produce subjective results, and the differences in conversation quality are not easily demonstrable through a simple question-and-answer format.
What is the purpose of the Pygmalion models?
-The Pygmalion models are specialized for conversation and role play, and they are not meant for general question answering. They excel in creating engaging and interactive dialogues, particularly in uncensored scenarios.
How does the script demonstrate the capabilities of the Pygmalion models?
-The script demonstrates the capabilities by having the creator engage in a role-play conversation with a character named Sandra, using both the 13B and 7B models to showcase their conversational abilities.
What tools are used to facilitate the conversation with the Pygmalion models?
-The creator uses the Uber Bugatti generation web UI with the Whisper extension from 11 Labs, allowing them to converse with the character using a microphone.
How can viewers access more characters created by the community for the Pygmalion models?
-Viewers can join the creator's Discord server or the Pygmalion AI Discord, or visit a website like guru.plus to find and download community-created characters.
What is the verdict on the quality of the conversations between the 13B and 7B models?
-The creator concludes that both models perform similarly, with the 7B model having slightly more action cues, which may make the scenario feel more realistic. However, the differences are subtle and may not be noticeable to all users.
What is the creator's final recommendation for users interested in the Pygmalion models?
-The creator encourages users to try the models themselves, especially when combined with written characters and the Whisper and 11 Labs extensions, as it can greatly enhance the role-play experience.
Outlines
🚀 Introduction to the New Pygmalion Model
The video begins with excitement over the release of the new Pygmalion 13 billion model, a significant upgrade from the previous 7 billion model. The creator discusses their initial reluctance to make another video about language models but decides to do so due to their fondness for the Pygmalion series. They mention the difficulty in visually demonstrating the differences between conversational models and highlight the new model's impressive performance in perplexity score benchmarks, outperforming other models like Stable Diffusion, DALL-E, and the original GPT-3.
🗣️ Comparing Conversational Models
The creator explains the challenge of comparing conversational models on camera due to the subjective nature of conversational results. They describe their testing of the new Pygmalion 13 billion model and its superior performance on paper. The video then transitions into a role-play scenario using the new model, where the creator interacts with a character named Sandra. The conversation revolves around daily life, work, and plans for the evening, showcasing the model's ability to generate coherent and contextually appropriate responses.
🍽️ Role-Play with Sandra
The video continues with a role-play conversation between the creator and the Sandra character, using both the new Pygmalion 13 billion model and the older 7 billion model. The creator compares the two models' performances, noting the 13 billion model's coherence and precision. They also discuss the limitations of the 7 billion model and the addition of action cues in the text, which enhances the realism of the scenario. The creator concludes that both models perform similarly well, but the 7 billion model has a slight edge in action cues. They encourage viewers to share their preferences in the comments.
Mindmap
Keywords
💡Pygmalion model
💡Perplexity score Benchmark
💡Conversational models
💡Role-play
💡Un censored
💡VRAM
💡Discord server
💡Environmental scientist
💡Netflix and chill
💡YouTube algorithm
Highlights
The new Pygmalion model has been released, an upgrade from the previous Bigvanion 7B model.
The Pygmalion model is designed for conversation and role-play, making it unsuitable for general question answering.
The Pygmalion 13B model performed better than several other 13B parameter models in the perplexity score benchmark.
The video creator is using the Uber Bugatti generation web UI with the Whisper extension for interactive role-play.
The creator's character, Sandra, is used to demonstrate the capabilities of the Pygmalion 30B and 7B models.
The conversation between Sandra and the creator is coherent and natural, showcasing the model's strength in role-play scenarios.
The creator discusses the environmental scientist character's work and personal life in a realistic and engaging manner.
The Pygmalion model's performance is compared to the 7B model, with the 13B model showing more detailed and precise responses.
The creator invites viewers to share their opinions on which model they prefer based on the conversation examples provided.
The video emphasizes the fun and immersive nature of using the Pygmalion models for role-play, especially when combined with other tools.
The creator thanks the viewers for their support and encourages them to try out the models themselves.
The video concludes with a call to action for viewers to subscribe, like, and support the creator's content.
The creator's Patreon supporters are acknowledged for their contribution to the production of the video content.
The video ends with a promise to see the viewers in the next video, maintaining engagement and anticipation.