Hunyuan Video LoRA Inference

Hunyuan Video LoRA Inference is centered on the tencent/hunyuan-video-lora workflow for hunyuan video lora work in video production. The endpoint is especially useful when a project has a defined transformation goal and needs a repeatable request contract instead of an ad-hoc desktop step. Teams can place this named operation inside review, asset preparation, and publishing pipelines while keeping the original request attached to every result. Hunyuan Video Lora by Tencent - generate cinematic videos from text descriptions with AI. Create high-quality video content with natural motion, camera control, and optional audio generation. Because tencent/hunyuan-video-lora is a separate catalog entry, evaluations should use this route’s own inputs, output expectations, and billing behavior. This is the creative remit of the exact endpoint, not a family-wide promise.

seed controls the seed to use for generating the video; loras controls the LoRAs to use for the image generation. You can use any number of LoRAs and they will be merged together to generate the final image; prompt controls the prompt to generate the video from; pro mode controls by default, generations are done with 35 steps. Pro mode does 55 steps which results in higher quality videos but will take more time and cost 2x more billing units; num frames controls the number of frames to generate; num frames accepts 129, 85; resolution controls the resolution of the video to generate; resolution accepts 480p, 580p, 720p. The mandatory portion is prompt. Hunyuan Video LoRA Inference exposes seed, loras, prompt, pro_mode, num_frames, resolution, aspect_ratio as its documented request surface. For example, the route documentation includes “A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage. She wears a black leather jacket, a long red dress, and black boots, and carries a black purse.”. Applications can submit that exact shape asynchronously, retain the chosen arguments in job metadata, and collect the resulting media URL for review or publishing.

Highlights

• Coherent motion. Hunyuan Video LoRA Inference is built to turn the route’s text or reference media into a temporally connected sequence rather than unrelated frames.
• Shot direction. Prompts can describe subject action, camera movement, pacing, and atmosphere so Hunyuan Video LoRA Inference has a concrete motion plan.
• Reference-aware generation. When this route accepts source media, Hunyuan Video LoRA Inference uses it to anchor identity, composition, or the clip boundary throughout generation.
• Production variants. Declared duration, resolution, framing, or inference controls support repeatable video options for review and selection.

Pricing

When to Use

Prompt Guide

Lead with the subject or transformation, then add the composition, environment, style, lighting, motion, voice, or fidelity details relevant to this route. Keep every parameter inside the documented schema; the example below uses only fields declared for tencent/hunyuan-video-lora.

{
  "prompt": "A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage. She wears a black leather jacket, a long red dress, and black boots, and carries a black purse.",
  "num_frames": "129",
  "resolution": "480p",
  "aspect_ratio": "21:9"
}

Technical Specs

Related Models

• tencent/hunyuan-motion
• tencent/hunyuan-portrait
• tencent/hunyuan-custom
• tencent/hunyuan-world