Playht

Play.ht provides one of the largest AI voice libraries (900+ voices, 142 languages) with a strong focus on ultra-realistic speech synthesis. PlayHT 3.0 model produces natural, expressive speech with emotional range. Instant voice cloning from a 5-second sample makes personalized voice generation accessible. For agents: generate voiceovers for content, power TTS in conversational AI pipelines, automate podcast production, or build multilingual voice interfaces. REST API with streaming support; lower latency than fully synchronous endpoints. Competitive with ElevenLabs for voice diversity and language coverage. Confidence is docs-derived.

REST API at api.play.ht/api/v2. Resources: tts, voices, cloned-voices. POST /api/v2/tts submits a TTS job ({text, voice, output_format: mp3/wav/ogg/flac/mulaw/alaw, quality, speed, sample_rate, voice_engine}). Returns a streaming response (SSE: server-sent events with audio_url). GET /api/v2/tts/{id} retrieves job status and audio URL. POST /api/v2/tts/stream streams audio bytes directly (response body is audio). GET /api/v2/voices lists all available AI voices with metadata (language, gender, style). POST /api/v2/cloned-voices/instant creates an instant voice clone (multipart audio upload, voice_name). Response: JSON (job responses) or audio bytes (stream endpoint).

Dual auth: Authorization: Bearer {secret_key} and X-USER-ID: {user_id} headers both required. Keys from Play.ht API Access page in dashboard. Account-specific user_id paired with secret key. HTTPS enforced. No OAuth2. Cloned voices account-scoped. Rate limiting: plan-dependent; standard limits on concurrent TTS jobs. Key rotation via Play.ht Dashboard.

Standard HTTP status codes with JSON error bodies. 401 for invalid credentials (both headers required — missing either returns 401). 400 for invalid voice ID, unsupported format, or text exceeding limits. TTS job: SSE stream emits events including audio_url when ready (for async flow) or audio bytes directly (for streaming flow). Queue time depends on plan and server load. Play.ht uptime at status.play.ht. Streaming is preferred over polling for latency-sensitive use cases. Audio file storage: generated files hosted on Play.ht CDN (time-limited download URLs).

docs.play.ht provides REST API reference for TTS, voice listing, and voice cloning endpoints. Python SDK (pyht) and Node.js SDK available. Getting started: create Play.ht account, access API keys in dashboard, first TTS request in under 10 minutes. Dual auth (both Authorization and X-USER-ID headers) is a minor gotcha documented in the quickstart. Voice browser in the Play.ht UI helps identify voice IDs before API integration. Community support via Play.ht Discord and support chat.