CapCut TTS / STT API

Project Overview

A pure Python command-line client that reverse-engineers CapCut's internal common-task API to expose Text-to-Speech (TTS) and Speech-to-Text / subtitle recognition (STT) as scriptable CLI commands. All request construction, RSA payload signing, AWS SigV4 authorization, and VOD upload are implemented in Python from scratch — zero native libraries, zero ctypes, zero .dylib dependencies. Designed for developers who need programmatic access to CapCut's high-quality TTS voices and ASR engine.

Key Features

• TTS: build SSML → sign RSA payload → POST to CapCut → poll task result → receive audio
• STT: upload audio/video via VOD (AWS SigV4) → submit ASR task → poll subtitles
• All-in-one stt-file command: upload + submit STT in a single CLI call
• RSA PKCS#1 v1.5 payload signing — pure Python, no openssl binary needed
• AWS SigV4 for ApplyUploadInner and CommitUploadInner — pure Python
• x-ss-stub (MD5 body hash) and request sign header generation
• --device-json override for custom CapCut device/session profiles
• --dry-run flag to preview signed request without calling the API
• Subtitle extractor: parses utterances with start/end timestamps and word-level data
• Open source, MIT-friendly — supports vi-VN and other CapCut locales