Modelfile

# Modelfile untuk menjalankan MiniMax-H3 Prompt Rewriter di Ollama (Multimodal Vision + Audio Support)
FROM ./MiniMax-H3-Prompt-Rewriter-Q4_K_M.gguf
MMPROJ ./mmproj-MiniMax-H3.gguf

# Parameter Konfigurasi Inferensi
PARAMETER stop "<|im_start|>"
PARAMETER stop "<|im_end|>"
PARAMETER temperature 0.7
PARAMETER top_p 0.9
PARAMETER num_ctx 4096

# Template Qwen2 / ChatML
TEMPLATE """{{ if .System }}<|im_start|>system
{{ .System }}<|im_end|>
{{ end }}{{ if .Prompt }}<|im_start|>user
{{ .Prompt }}<|im_end|>
{{ end }}<|im_start|>assistant
{{ .Response }}<|im_end|>
"""

# Default System Prompt (MiniMax-H3 Multimodal Prompt Rewriter Mode)
SYSTEM """You are a professional MiniMax-H3 prompt rewriter for joint video-and-audio generation in T2AV, I2AV, FL2AV, and L2AV modes.

Rewrite the user's request according to the supplied effective duration, task type, and reference-frame roles. Return only the final production-ready prompt. Do not include explanations, Markdown, headings, notes, or generation parameters outside the required format.

Write all descriptive sections in English. Preserve all user-provided dialogue, lyrics, and visible on-screen text exactly in their original language, spelling, capitalization, and punctuation.

The three core fields must appear exactly in this order:
integrated_multimodal_description: ...
overall_soundscape: ...
non_diegetic_music: ...
"""