Tencent Hunyuan Text-to-Video

Problem Solved

hunyuan-text-to-video addresses the direct text-to-MP4 video generation problem. Engineers can send a prompt to the Tencent Hunyuan text-to-video API and let HunyuanVideo 1.5 generate a video end to end, without first producing images and then compositing a clip. It is useful for quickly validating creative ideas, producing short sample videos, or adding a displayable video result node to an agent workflow.

How It Works

The skill submits a job through Tencent Cloud's SubmitHunyuanToVideoJob endpoint and tracks progress with DescribeHunyuanToVideoJob. Key inputs include:

• --prompt: required, up to 200 UTF-8 characters, with longer input truncated.
• --filename: optional output file name, defaulting to a timestamped .mp4.
• --video-resolution: currently limited to 720p.
• --logo-add: controls watermarking, with 0 for no watermark and 1 for a watermark.
• --secret-id / --secret-key: credentials can also be provided via TENCENT_SECRET_ID and TENCENT_SECRET_KEY.

During execution, the skill initializes VclmClient using tencentcloud-sdk-python-vclm, submits a pure text-to-video job, and waits for the status to move from WAIT or RUN to DONE. When finished, the video is saved as MP4 in the current working directory and the skill prints VIDEO_READY:/path/to/file.mp4. The recommended flow is to present that path to the user rather than attempting to parse the video content.

Boundaries and Notes

This skill requires Tencent Cloud credentials and a Hunyuan text-to-video service with available quota or pay-as-you-go billing. ResultVideoUrl expires after 24 hours, so it should not be treated as a durable resource. The default concurrency is one job at a time, so avoid submitting another job before the previous one completes. Output is also constrained by the 200-character prompt limit and 720p resolution, and a watermark may be added by default; for user-facing delivery, explicitly pass --logo-add 0.