Venice AI Media

**Judgement:** Strong, feature-rich media generation skill with broad coverage of image and video workflows, but with notable safety concerns due to default "uncensored" behavior. **What it does well** - Provides end-to-end support for text-to-image, upscaling, image editing, and image-to-video, all via simple Python CLI scripts. - Exposes rich controls (models, resolution, CFG, seeds, styles, LoRA strength, video duration/resolution, audio options), making it suitable for power users. - Integrates cleanly with Clawdbot through MEDIA-tagged outputs and supports persistent API key configuration. - Includes practical operational features like `--quote` for video cost estimation and `--list-models`/`--list-styles` for discoverability. **Key risks / limitations** - Venice is explicitly described as an uncensored API, and `safe-mode` is **disabled by default**. This significantly increases the risk of generating NSFW or otherwise sensitive content if prompts are not carefully constrained. - Models can change frequently; reliance on `--no-validate` and evolving model lists can introduce occasional breakage or unexpected behavior. - Reliability and performance are tied to Venice’s external service quality (latency, uptime, model availability), which is outside the skill’s control. **Recommended scenarios** - Best suited for technically comfortable users (or automated agents) who need powerful, flexible media generation and are operating in environments where uncensored outputs are acceptable and properly governed. - Appropriate for creative prototyping, content generation for marketing or social media, and experimentation with various image/video models. - Should be used with **explicit prompt controls and, where needed, `--safe-mode`** in any context that requires content moderation, compliance, or brand safety.