Sample code showing how to call Alibaba Cloud Bailian speech models for speech recognition, speech synthesis, and voice chat.
42958MITUpdated 5 days ago
PedroMarinhoDev/ComfyUI-Drama3
ComfyUI nodes for Fish Audio Drama 3 text-to-speech with voice cloning, multi-speaker dialogue and inline cues, plus native Qwen3-ASR transcription.
- Stars
- 165
- Forks
- 21
- Open issues
- 6
- Last commit
- 14 days ago
What is it?
ComfyUI-Drama3 brings Fish Audio's Drama 3 text-to-speech model into ComfyUI as a set of nodes. It supports cloning a voice from a short clip and its transcript, saving that voice to reuse later, generating dialogue for up to 16 speakers, and inserting cues such as [laughs], [whispers] or [sighs] into the text. Long text is split into chunks and generated in sequence.
The model is a 35B mixture-of-experts network. Most of its weights are routed experts that are kept in system RAM, while a VRAM cache holds the experts in active use, which the project says allows it to run on GPUs from 8 GB. The pack also includes a native Qwen3-ASR node that transcribes audio, which can fill in a voice clip's transcript automatically. Model weights are downloaded separately from Hugging Face and are licensed under the Fish Audio Community License, while the code is GPL-3.0.
Key features
- 01Clones a voice from a short audio clip and its transcript, and saves it as a small reusable file.
- 02Generates dialogue for up to 16 speakers, each with its own voice.
- 03Inline cues such as [laughs], [whispers] and [sighs], plus scoped tags for pacing and emphasis.
- 04Keeps expert weights in system RAM with a VRAM cache to run on cards from 8 GB.
- 05Includes native Qwen3-ASR speech recognition in 30 languages.
- 06Editor node provides word-level captions and per-voice tracks.
Use cases
- Producing voiceover or narration from a script inside a ComfyUI workflow.
- Prototyping multi-character dialogue for audio drama or animation.
- Transcribing voice clips automatically to prepare them for cloning.
- Running a local text-to-speech pipeline on a consumer GPU.
Quick start
- 1
Install the nodes with ComfyUI-Manager or the Comfy Registry
comfy node install drama3 - 2
Clone the repo into custom_nodes and restart ComfyUI
cd ComfyUI/custom_nodes git clone https://github.com/PedroMarinhoDev/ComfyUI-Drama3 - 3
Download the model weights from Hugging Face
hf download PedroMarinhoDev/Drama-3-ComfyUI drama3_fp8experts_bf16.safetensors --local-dir ComfyUI/models/diffusion_models - 4
Open the example voice clone workflow and queue it in ComfyUI
Things to know
- Requires an NVIDIA GPU with at least 8 GB VRAM and roughly 32 to 48 GB of system RAM.
- The model files are very large, about 38.6 GB for the default FP8 file.
- Code is GPL-3.0 while the model weights use the separate Fish Audio Community License.
In collections
Topics
#text-to-speech#voice-cloning#comfyui#speech-recognition#audio#asr#speech-synthesis#comfyui-nodes#fish-audio
FAQ
- What is ComfyUI-Drama3?
- ComfyUI nodes for Fish Audio Drama 3 text-to-speech with voice cloning, multi-speaker dialogue and inline cues, plus native Qwen3-ASR transcription. ComfyUI-Drama3 brings Fish Audio's Drama 3 text-to-speech model into ComfyUI as a set of nodes. It supports cloning a voice from a short clip and its transcript, saving that voice to reuse later, generating dialogue for up to 16 speakers, and inserting cues such as [laughs], [whispers] or [sighs] into the text. Long text is split into chunks and generated in sequence.
- Is ComfyUI-Drama3 free to use? What license does it have?
- ComfyUI-Drama3 is open source under the GPL-3.0 license. Check the license terms before using it in commercial products.
- Is ComfyUI-Drama3 still maintained?
- Actively maintained. The last commit to ComfyUI-Drama3 was on Oct 25, 2026.
- What can I use ComfyUI-Drama3 for?
- Producing voiceover or narration from a script inside a ComfyUI workflow. Prototyping multi-character dialogue for audio drama or animation. Transcribing voice clips automatically to prepare them for cloning. Running a local text-to-speech pipeline on a consumer GPU.
- How do I install ComfyUI-Drama3?
- Install the nodes with ComfyUI-Manager or the Comfy Registry: comfy node install drama3 Clone the repo into custom_nodes and restart ComfyUI: cd ComfyUI/custom_nodes git clone https://github.com/PedroMarinhoDev/ComfyUI-Drama3 Download the model weights from Hugging Face: hf download PedroMarinhoDev/Drama-3-ComfyUI drama3_fp8experts_bf16.safetensors --local-dir ComfyUI/models/diffusion_models Open the example voice clone workflow and queue it in ComfyUI
- Where can I get ComfyUI-Drama3?
- The source code and installation instructions are on GitHub: https://github.com/PedroMarinhoDev/ComfyUI-Drama3
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A native ComfyUI custom node that brings the Ming-3.0-tts speech generation models (0.5B and Tiny 9B) into ComfyUI workflows.
28+1 this week2PythonMITUpdated 8 days ago
Open-source voice AI family from Microsoft, with long-form speech recognition (ASR), multi-speaker text-to-speech (TTS), and real-time streaming TTS models.
56.7K+1.6K this week6.4KPythonMITUpdated last month
A C++ speech toolkit on ggml for text-to-speech, speech recognition and voice activity detection, running on Metal, Vulkan or CPU with an OpenAI-compatible server.
AI & Machine Learning#speech-synthesis#speech-recognition#text-to-speech#voice-activity-detection#ggml
371+93 this week23C++MITWell maintainedUpdated 6 days ago
A Python library for text-to-speech, speech-to-text and speech-to-speech models running natively on Apple Silicon via the MLX framework.
8.2K+46 this week787PythonMITWell maintainedUpdated 23 hours ago
VoxCPM2 is an open-source, tokenizer-free text-to-speech model supporting 30 languages, voice design from text descriptions, and voice cloning from short reference clips.
41.3K+364 this week4.7KPythonApache-2.0Well maintainedUpdated 3 days ago