ComfyUI Integrates HappyHorse 1.1 and Krea 2 as Local Video Workflows Expand

TL;DR

ComfyUI added support for HappyHorse 1.1 and Krea 2 this week, enabling filmmakers to run audio-native video generation and high-fidelity image models locally (ComfyUI Blog). For AI filmmakers, this consolidation reduces the need for multiple disparate platforms by housing video, dialogue, and sound effects within a single node-based environment.

What Happened

ComfyUI expanded its ecosystem by integrating HappyHorse 1.1, a model designed to generate synchronized video, dialogue, and sound effects simultaneously (ComfyUI Blog). This update allows creators to bypass the traditional step of manual audio-to-video alignment, as the model handles temporal synchronization within the generation process itself.

Parallel to this, Krea 2 released open weights for its RAW and Turbo models, which are now accessible via ComfyUI (Krea Blog). The RAW model focuses on photographic accuracy and texture, while the Turbo model is optimized for real-time generation. This release provides a local alternative to cloud-based image generators, giving filmmakers more control over their hardware and data privacy (ComfyUI Blog).

Further broadening the platform's utility, Seedance 2.0 Mini and a 4K high-resolution model were added to ComfyUI for advanced video-to-video tasks (ComfyUI Blog). These models specifically target character consistency, a persistent challenge in AI cinematography. Additionally, ComfyUI launched a native mobile application framework for Comfy Cloud, allowing users to execute node-based workflows on iOS and Android devices via a public API (ComfyUI Blog).

Why This Matters

Before these integrations, generating a cohesive scene required jumping between a video generator, a voice cloner, and a foley tool. The HappyHorse 1.1 integration in ComfyUI changes this by treating audio as a native component of the video file from the start. This reduces the "uncanny valley" effect often caused by slightly misaligned lip-sync or sound effects that do not match the visual physics of a scene.

Krea 2's move to open weights is equally significant for the production pipeline. By making the RAW and Turbo models available locally, Krea allows editors to generate high-fidelity assets without recurring credit costs or cloud latency. For directors, the ability to use ComfyUI to maintain style consistency across hand-drawn and 3D assets, as seen in the Oscar-shortlisted film The Little Poet, demonstrates that these tools are moving beyond experimental clips into professional-grade short-form cinema (ComfyUI Blog).

For AI Filmmakers

Directors and VFX artists can now build more complex, automated pipelines that handle everything from character turnarounds to final sound-mixed exports. If you are working on character-heavy narratives, the Seedance 2.0 Mini update is the primary tool for maintaining visual identity across multiple shots. For those struggling to describe complex visual styles for the Krea RAW model, the Image to Prompt tool can help reverse-engineer the necessary descriptors from reference frames.

Storyboarding also becomes more efficient with these updates. Instead of static images, the integration of HappyHorse 1.1 allows for "living storyboards" where the pacing of dialogue is baked into the initial pre-visualization. Creators can use the AI Film Storyboard toolkit to organize these prompts before moving them into the ComfyUI node environment for final generation.

What To Do Now

1. Update your ComfyUI installation to the latest version to access the new HappyHorse 1.1 and Krea 2 nodes.

2. Download the Krea 2 RAW weights if your project requires high-fidelity textures for close-up shots.

3. Test the Seedance 2.0 Mini model for video-to-video tasks where maintaining the actor's face and clothing is critical.

4. Experiment with the Comfy Cloud mobile API if you need to monitor or trigger long render queues while away from your workstation.

The Bigger Picture

The trend toward local, open-weight models like Krea 2 and unified audio-video models like HappyHorse 1.1 suggests that the AI film industry is moving away from fragmented web-apps. ComfyUI is positioning itself as the central operating system for these models, mirroring the way DaVinci Resolve or Premiere Pro consolidated editing, color, and sound in previous decades. As Google also moves toward "computer use" capabilities in Veo 3 (Google DeepMind), the line between the AI model and the creative software interface continues to blur.

Sources & further reading click to expand

• ComfyUI Blog: HappyHorse 1.1 Integration
• ComfyUI Blog: Krea 2 RAW and Turbo Models
• Krea Blog: Krea 2 Technical Report
• ComfyUI Blog: Seedance 2.0 Mini and 4K Support
• ComfyUI Blog: Case Study: The Little Poet
• ComfyUI Blog: Comfy Cloud Mobile App
• Google DeepMind: Computer Use in Gemini 3.5 Flash