Speech

Generates spoken audio from text via the OpenAI Audio API. You get a bundled CLI that handles single clips or batch jobs (JSONL), defaults to gpt-4o-mini-tts with built-in voices like cedar and marin, and supports instructions for pacing, tone, and emphasis. The workflow is opinionated: collect exact text up front, augment into a labeled spec without rewriting, validate the output, iterate with single targeted changes. Use it for narration, product demos, IVR prompts, or accessibility reads. The references folder gives you templates for each use case, instruction patterns, and API limits. Custom voices are out of scope. Requires OPENAI_API_KEY and respects the 50 RPM rate limit.