fish-speech

குரல் மாதிரி மட்டும் போதும்.. Fish Speech அதே குரலில் உணர்ச்சியோடு பேச வைக்கிறது. நூறு மொழிகள் வரை இயங்குகிறது. முதலில் எதைப் பேச வைப்பீர்கள்? fishaudio/fish-speech Instagram-லும் பின்தொடருங்கள்: instagram.com

@SiamKidd The one I mentioned, Fish fishaudio/fish-speech

Fish Audio가 오픈소스 음성 모델 Drama 3를 공개했습니다. 35B MoE 기반으로 자연어 연기 디렉팅을 지원합니다. 감정 태그로 톤을 조절하며 다회차 피드백으로 수정할 수 있습니다. 로컬은 FP8 기준 48GB VRAM이 필요하고, 웹에서 매달 8,000 크레딧을 무료 제공합니다. fishaudio/fish-speech t.co

Fish Speech — FREE Open Source TTS with Real EMOTION (Full Tutorial) 🤖 Tired of robotic text-to-speech? Meet Fish Speech — the FREE open-source TTS (33K+ GitHub stars) that doesn't just read your text, it PERFORMS it. In this tutorial: ✅ Emotion tags — make it whisper, laugh, or sound angry on command ✅ Voice cloning from just 10 seconds of audio ✅ Multi-speaker dialogue in a single pass ✅ Try it free in your browser (no install needed) INSTALL: git clone fishaudio/fish-speech cd fish-speech pip install -U "huggingface_hub[cli]" hf download fishaudio/s2-pro --local-dir checkpoints/s2-pro docker run --gpus all -v ./checkpoints:/app/checkpoints -p 8080:8080 fishaudio/fish-speech:server-cuda Then open localhost Try it: fish.audio Code: github.com/fishaudio/fish-speech #OpenSource #AI #TextToSpeech #VoiceCloning #Tutorial #FreeAI #Tech #AltaMedia

Fish Audio Drama 3 emotional dialogue test (open weights) Tried Fish Audio Drama 3 on a short scene between two people saying goodbye. I wanted to test a few shifts within the same conversation: anger, a quiet plea, then trying to keep it together. The “Don't. Please.” line was the part I wanted to push most. I used inline text directions for caught breaths, hesitation, a quieter voice, and trying to speak through tears. You can place these partway through a line to change the delivery as the character speaks. The captions show the full directions used for each line. Drama 3 is an open-weight TTS model released under the Fish Audio Community License. Model / weights: huggingface.co GitHub: fishaudio/fish-speech Docs: docs.fish.audio

声優を雇わなくていい時代が来た。 fish-speechで10秒の音声サンプルから話者の声を完全再現できる。 ナレーション制作コストがゼロになる。 fishaudio/fish-speech teikoku-signal.netlify.app

Fish Audio Releases Drama 3: a 35B Open-Weight MoE Text-to-Speech Model Fish Audio just released Drama 3, an open-weight text-to-speech model that takes stage directions in plain language. It is a 35B MoE with about 3B active parameters per token, covering 83 languages and up to 16 speakers in a single request. The core innovation is a hybrid backbone paired with a new Codebook-MTP decoder. 3 Gated DeltaNet layers sit between each full-attention layer, so only 1 layer in 4 keeps a KV cache. Post-training uses GRPO with a single learned human-preference reward model. Drama 3 scores 93.2% on Artificial Analysis Pronunciation Robustness, up from 88.1% for S2.1 Pro. The trade-off: its quality Elo of 1242 still trails Eleven v4 Turbo at 1328, and exact sequences like codes and emails lag at 86.3%. Self-hosting needs an NVIDIA GPU with 48 GB+ in FP8, and free commercial use stops at $500K in annual revenue...... Full analysis: t.co Model: t.co Technical report: t.co

Fish Audio Drama 3 Technical Report We introduce Fish Audio Drama 3, an open-source speech generation model that turns text into directed performances, built on a 35B-parameter mixture-of-experts backbone that activates 3B parameters per token. It clones voices from multiple references, follows natural-language direction, revises a take when the next turn gives a note, voices multi-character scenes, edits existing speech by infilling, and synthesizes long-form and streaming audio. Multi-turn revision, scene dubbing, editing by infilling, and streaming text input are new in Drama 3, all in one model. Drama 3 pairs the hybrid linear-attention backbone with a new codebook multi-token-prediction audio decoder that predicts the acoustic codebooks of each frame in dedicated layers that attend over the entire sequence. The model is trained with a three-stage pre-training recipe, instruction fine-tuning, and large-batch reinforcement learning whose only reward is a single learned model of human preference. Its inference engine decodes a single stream at 271 frames per second on one H200, against 177 for Fish Audio S2, although Drama 3 has more than seven times as many parameters. The model weights are available on GitHub (fishaudio/fish-speech) and Hugging Face (huggingface.co), and inference is available through SGLang-Omni (sgl-project/sglang-omni). We highly encourage readers to visit fish.audio to direct their own performances.

做配音、有声书，开源语音方案已经够用了。 Fish Speech 中文效果好，一段几十秒的样本就能克隆音色，本地跑不按字数付费。 fishaudio/fish-speech