Loading
Vous avez un workflow ComfyUI ? Obtenez une documentation lisible directement dans votre navigateur.Rendre un workflow lisible.Essayer maintenant ↗
A more advanced interaction thanks to a new continuity mode: the last frame of the video is isolated and fed to the model.
Profil d'exécution
Description de la source
A more advanced interaction thanks to a new continuity mode: the last frame of the video is isolated and fed to the model.
This is not by any means a product but a concept-prototype.
It uses ollama, but if you don't like ollama you probably can figure how to change that yourself. My experience with LLMs is limited.
This is how you play the game:
First round. An image, a setting, a voice, and two prompts one for the LLM and one that only Minimax will read. There are two toggles Continuity and Reset Memory. Continuity to false and Reset Memory to True. Make sure that the Continuity nodes can read an image even if it will not be used.
Hit run and wait for the round to complete. In my 4060 Ti with the backup of a macbook M1 executing the LLM every round is about 4 minutes.
Commentaire généré par l’IA
Explication générée par l’IA à partir des détails de la source et de la configuration. Les suggestions sont clairement signalées.
Companion Engine is a ComfyUI workflow for spoken avatar video replies: give it an avatar image, a setting, a voice sample, a question, and prompts, and it can create a video answer that carries its last frame into the next run.
Ollama sends the question and avatar image to a vision model; its JSON reply supplies a scene description and duration, which MiniMax H3 uses to render the speaking avatar with audio from the voice sample.
With continuity enabled, the last frame of the video is isolated and fed into the model for the next run.
Canvas controls include the avatar image, voice sample, Ollama server URL and model, question, resolution, and loaders for the diffusion model, text encoder, video VAE, and audio VAE.
The outputs are the video, the LLM's raw JSON, and the exact prompt sent to MiniMax.
Version v2.0 is listed for MiniMax H3, and the default Ollama server address is 127.0.0.1:11434.
The listed node packs are comfyui-easy-use, comfyui-impact-pack, and comfyui-ollama.
Listed model files are minimax_h3_audio_vae_fp32.safetensors, minimax_h3_ref2v_lightx2v_turbo_4step_v0.1_resized_avg_rank_20_bf16.safetensors, minimax_h3_ref2va_pruned_int8_convrot.safetensors, minimax_h3_video_vae_fp16.safetensors, and qwen3vl_32b_minimax_h3_nvfp4_awq.safetensor
For the first round, set Continuity off and Reset Memory on; for later rounds, turn Continuity on and Reset Memory off, then edit the conversation prompt.
Use the conversation prompt to describe the avatar's identity, personality, and goals or ask questions; a separate hidden prompt can guide content the companion may not know unless it sees it in a later frame.
This is described as a concept prototype, not a finished product.
No download location is listed for minimax_h3_ref2v_lightx2v_turbo_4step_v0.1_resized_avg_rank_20_bf16.safetensors, so check that you can obtain it before setup.
Suggestion · non vérifié
Faut-il modifier cela ?
Connectez-vous pour envoyer une demande de modification.
Sources
1 sourceExtraits de sources
1 extraitSource context: 333 downloads · Type Workflows · Base model MiniMax H3
Second round and on: change both toggles (Continuity true, Reset Memory false). Edit the conversation prompt and hit run.
You can experiment with the prompt, I use the prompt for the LLM to tell the avatar who she is, her personality and goals, and my questions. The hidden prompt you can use to create things that the companion isn't aware of, unless it sees it in the last frame when the next round comes.
Exigences
minimax_h3_audio_vae_fp32.safetensors
Fichier non vérifiéVAE · Hugging Face · Comfy-Org/MiniMax-H3 · MiniMax H3
minimax_h3_ref2v_lightx2v_turbo_4step_v0.1_resized_avg_rank_20_bf16.safetensors
Non résoluLoRA · Unknown
minimax_h3_ref2va_pruned_int8_convrot.safetensors
Fichier non vérifiéCheckpoint · Hugging Face · Comfy-Org/MiniMax-H3 · MiniMax H3
minimax_h3_video_vae_fp16.safetensors
Fichier non vérifiéVAE · Hugging Face · Comfy-Org/MiniMax-H3 · MiniMax H3
qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors
Fichier non vérifiéText encoder · Hugging Face · Comfy-Org/MiniMax-H3 · MiniMax H3
comfyui-easy-use
TrouvéPack de nœud · Registry
comfyui-impact-pack
TrouvéPack de nœud · Registry
Pack de nœud · Registry
