The recommended model strength is 0.7; for segments with high motion, it is advisable to lower this to 0.5. Trigger word: DY.
Online Workflow Experience: MiniMax H3 – Authentic Cinematic Quality
Try it here: https://www.runninghub.ai/zh-cn/post/2092446042583642113/?inviteCode=rh-v1635
Sign up to receive 1,000 credits and try it for free.
This is a workflow designed to achieve a cinematic look with MiniMax H3, specifically optimized to address common AI video issues such as a "plastic" appearance, over-sharpening, harsh lighting, and an artificial "AI feel." By integrating a specially trained LoRA for authentic cinematic texture—while preserving MiniMax H3’s inherent strengths in motion, shot comprehension, and audiovisual generation—the workflow enhances natural lighting, color depth, depth of field, skin texture, atmospheric mood, and lens characteristics. The result is an output that closely resembles genuine cinematography rather than merely applying a cinematic filter. It is well-suited for narrative shorts, dialogue-driven scenes, cinematic advertisements, and genres such as martial arts, suspense, and urban drama, as well as for capturing evocative, mood-driven shots; it also serves as a versatile solution for adding a cinematic quality to everyday video generation with H3.
Description
The recommended model strength is 0.7; for segments with high motion, it is advisable to lower this to 0.5. Trigger word: DY.
Online Workflow Experience: MiniMax H3 – Authentic Cinematic Quality
Try it here: https://www.runninghub.ai/zh-cn/post/2092446042583642113/?inviteCode=rh-v1635
Sign up to receive 1,000 credits and try it for free.
This is a workflow designed to achieve a cinematic look with MiniMax H3, specifically optimized to address common AI video issues such as a "plastic" appearance, over-sharpening, harsh lighting, and an artificial "AI feel." By integrating a specially trained LoRA for authentic cinematic texture—while preserving MiniMax H3’s inherent strengths in motion, shot comprehension, and audiovisual generation—the workflow enhances natural lighting, color depth, depth of field, skin texture, atmospheric mood, and lens characteristics. The result is an output that closely resembles genuine cinematography rather than merely applying a cinematic filter. It is well-suited for narrative shorts, dialogue-driven scenes, cinematic advertisements, and genres such as martial arts, suspense, and urban drama, as well as for capturing evocative, mood-driven shots; it also serves as a versatile solution for adding a cinematic quality to everyday video generation with H3.
FAQ
Comments (16)
does it influence sound and specifically voices?
no
Since no example prompts were given, here's one from the workflow they link to:
----------
DY
Macro close-up of a sleek foundation bottle resting on intricately textured, warm amber rock formations. A narrow, intense cinematic spotlight slowly sweeps across the scene from left to right in extreme slow motion. As the light moves, it elegantly reveals the smooth matte finish of the bottle and highlights the deep, rich crevices of the canyon-like rocks, creating a dramatic interplay of light and shadow. High-end commercial product cinematography, luxurious atmosphere, warm golden lighting, photorealistic, 8k resolution, smooth subtle camera push-in.
Hello, what's the difference between this lora and the other cinematic lora you posted a couple of days? Is it an update? Should I use both together?
They have the same trigger word. Here's a directly comparable output, though I have a hard time telling the difference, especially since the resolution is different, so it's hard to tell what comes from the extra resolution:
Authentic Cinematic Texture
https://civitai.red/images/140870189
Cinematic Look:
https://civitai.red/images/141591868
主要是解决张量报错的问题
@Jellai I see it similar aswell
@TuTu_1018 so only one lora is required? no need to activate both?
yes
can it use for ref2va ?
style lora works on ref2v but at 0.6-0.7 max. At 0.7+ break identity of your characters
@gambikules858 thanks
Great! Thank you for the update. Although the previous error didn't affect usage, it output thousands or even tens of thousands of lines of error messages to the console every time a video was generated, which slowed down my computer. I hope this issue can be resolved now.
yes,Resolved
给图图打CALL！
Imma wait for the (Full of Tensor Errors) version.