Model Overview
| Model | Model ID | Type | Features |
|---|---|---|---|
| GLM-TTS | glm-tts |
TTS | 7 voices, WAV/PCM, 24kHz |
| GLM-TTS-Clone | glm-tts-clone |
Voice Cloning | Clone with a 3-second sample |
| GLM-ASR-2512 | glm-asr-2512 |
ASR | CER 0.0717, supports dialects |
| GLM-4-Voice | glm-4-voice |
End-to-End Voice | 8K context, emotion/tone/speed control |
| GLM-Realtime | glm-realtime |
Real-Time Dialogue | WebSocket, supports interruption/function calling |
| GLM-Realtime-Flash | glm-realtime-flash |
Real-Time Dialogue | 0.18 CNY/minute |
| GLM-Realtime-Air | glm-realtime-air |
Real-Time Dialogue | 0.30 CNY/minute (including video) |
Selection Recommendations
- Standard TTS:
glm-tts - Personalized Voice:
glm-tts-clone— 3-second sample cloning - Speech-to-Text:
glm-asr-2512 - Voice Dialogue:
glm-4-voice— emotion/speed control - Real-Time Interaction:
glm-realtime— WebSocket, supports interruption
API Key Configuration
Load ~/.glm/.env from ZHIPU_API_KEY. Key acquisition:https://open.bigmodel.cn/usercenter/apikeys
GLM-TTS Features
- 7 preset voices
- WAV/PCM output, 24kHz sample rate
- First-frame latency ≤400ms
GLM-TTS-Clone Features
- Voice cloning with a 3-second audio sample
- API:
POST https://open.bigmodel.cn/api/paas/v4/voice/clone
GLM-ASR-2512 Features
- Maximum 25MB, maximum 30s
- CER 0.0717
- Supports Chinese dialect recognition
GLM-4-Voice Features
- 8K token context, 4K maximum output
- WAV/Base64 mono 16-bit 44100Hz
- Emotion, tone, speed, and dialect control
GLM-Realtime Features
- WebSocket protocol
- Supports: interruption, function calling, singing, screen sharing
- Flash: 0.18/1.20 CNY/minute; Air: 0.30/2.10 CNY/minute
Recommended Scenarios
TTS speech synthesis, personalized voice cloning, speech-to-text transcription, end-to-end voice assistant, real-time audio-video dialogue