Zhipu BigModel Free Text-to-Image and Text-to-Video

Problem

When prompts need to become usable cover art, social media images, e-commerce visuals, or short videos, calling the raw model API directly can be tedious because it involves API keys, model selection, batching, and result validation. This skill wraps Zhipu AI BigModel text-to-image and text-to-video workflows for task-oriented use, which is useful for quickly validating output before producing final assets.

Core capabilities and workflow

• Image generation: supports single-image and batch generation, with options such as cogview-3-flash, cogview-4-250304, and cogview-4 for different speed and quality tradeoffs.
• Video generation: supports cogvideox-flash, cogvideox-2, and cogvideox-3, and can generate videos with AI audio.
• Key steps: set BIGMODEL_API_KEY, choose a model and size such as 1024x1024, 1024x1792, or 1792x1024, then run the script or Python API to produce results.
• Prompting and parameters: include subject, style, lighting, atmosphere, and quality requirements; use concurrency controls for batch jobs to reduce interruptions.

Scope and caveats

It is best suited for teams that already have access to the Zhipu BigModel API and need image or short-video assets. flash models are useful for fast tests, while cogview-4 and cogvideox-3 are better for final output. If the prompt reports a missing BIGMODEL_API_KEY, configure the environment variable first. If video generation is unstable, simplify the prompt, switch models, or lower concurrency. For higher quality, combine --quality hd with a higher-quality model.