cstr/fireredtts3-GGUF

FireRedTTS3 GGUF (CrispASR)

GGUF conversion of FireRedTeam/FireRedTTS3 (Apache-2.0) for CrispASR.

• fireredtts3-base-f16.gguf — Qwen3-1.7B LLM backbone + PatchEncoder + DiT flow head + tokenizer (+ baked default English voice prompt)
• fireredtts3-redae-f16.gguf — RedAE continuous latent autoencoder (24 kHz, 25 Hz 64-d latents) + CAM++ speaker encoder

Zero-shot voice-cloning TTS, 24 languages + 21 Chinese dialects. All credit for the model goes to FireRedTeam (Xiaohongshu).