Bulle d'Encre - A preview model for comic generation
2 min readupdated
This model is a single ComfyUI checkpoint file combining base MOSS-Image versions with two built-in LoRA adapters. The model specializes in creating images in the style of European comics with anime motifs, panels, speech bubbles, and readable text. It is built on the MOSS-Image architecture and designed for those who want to create stylized illustrations and comic pages without connecting additional weights.
Key features
- Distinctive style: A blend of European comic art and anime aesthetics.
- Text handling: Support for generating readable captions and speech bubbles directly on images.
- Turbo integration: A half-baked acceleration LoRA that allows image generation in 15 steps.
- Ease of use: Requires no separate Turbo or style LoRAs.
- Compression options: Versions available in BF16 and optimized int8 ConvRot format.
What you can do
- Create multi-panel comic pages with story dialogue and captions.
- Generate scenes featuring two interacting characters with detailed clothing.
- Draw stylized album or book covers with customizable lettering.
- Experiment with long text prompts to get detailed illustrations.
Example
Recommended ComfyUI settings: resolution 1344×1792, 15 steps, CFG 2.5, Euler sampler with sgm_uniform scheduler, and flow shift 5.
Read more
DetailsResolution: Optimized for 1344×1792 pixels.
System requirements: Available in BF16 (10.16 GB) and lightweight int8 ConvRot (5.41 GB) versions.
Limitations: Complex scenes with two characters may require additional prompt iterations or seed tuning.
License: Apache 2.0 (based on OpenMOSS).










Comments