nabin2004/qwen-Manimator-1-merged

TEXT GENERATIONPricing:Input $0.468 / Output $1.82Concurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 17, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

nabin2004/qwen-Manimator-1-merged is an 8 billion parameter Qwen3-based language model fine-tuned by nabin2004. It specializes in generating complete, executable ManimCE and Manim Voiceover animations, including audio bookmarking for precise audio-visual synchronization. This model excels at producing pedagogically rich Manim scripts with a 'Plan-then-Code' reasoning block, making it ideal for educational content creation.

Loading preview...

Overview

qwen-Manimator-1-merged is an 8 billion parameter model based on Qwen/Qwen3-8B, fine-tuned by nabin2004. Its primary purpose is to generate high-quality, pedagogically rich ManimCE and Manim Voiceover animations. The model integrates advanced features for creating synchronized audio-visual educational content, making it a specialized tool for educators and content creators.

Key Capabilities

  • VoiceoverScene Architecture: Generates full Manim scripts that inherit from VoiceoverScene, ensuring compatibility with Manim Voiceover.
  • Audio Bookmarking: Automatically includes <bookmark mark='NAME'/> tags and self.wait_until_bookmark("NAME") for precise audio-visual synchronization within animations.
  • CE API Compliance: Adheres strictly to Manim Community Edition syntax, avoiding deprecated legacy APIs for robust and future-proof code.
  • Plan-then-Code: Outputs a <Plan> reasoning block before the Python implementation, providing a clear thought process for the generated code.

Training Details

The model was fine-tuned using the nabin2004/qwen-Manimator-1-sft-data dataset over 3 epochs, with a maximum sequence length of 4500 tokens. It utilized a LoRA configuration with r=16 and alpha=32, and an adamw_8bit optimizer.

Good For

  • Automated Educational Content Creation: Ideal for generating complex mathematical and scientific animations with integrated voiceovers.
  • Manim Script Generation: Developers and educators looking to quickly produce ManimCE scripts that are ready for execution.
  • Audio-Visual Synchronization: Users requiring precise timing between visual elements and audio narration in their Manim projects.

Finetunes
1 models