| OGG TTS for ComfyUI: licensing of what this pack contains | |
| The pack as a whole is distributed under the GNU General Public License v3 (LICENSE), because it includes code taken from | |
| ComfyUI-Drama3, which is GPL-3.0. Parts keep their own licenses and notices: | |
| 1. OGG TTS (https://github.com/FashionFlora/ogg-tts, Apache-2.0, Copyright 2026 FashionFlora; LICENSE-APACHE) | |
| ogg/model.py is ogg_tts/omnifish.py; ogg/generate.py holds prefix_ids, voice_context, _Stream, blocks_cached and speech_frames of | |
| ogg_tts/infer_omnifish.py; ogg/id_vocab.json is ogg_tts/id_vocab.json. | |
| Changes: training-machine paths removed; the MTP's prompt cross-attention switch is a constructor argument chosen from the | |
| checkpoint's keys; from_omnivoice became from_config_dir (config and tokenizer only); the CLI, streaming and checkpoint loading | |
| were left out. | |
| 2. ComfyUI-Drama3 (https://github.com/PedroMarinhoDev/ComfyUI-Drama3, GPL-3.0, Copyright PedroMarinhoDev) | |
| ogg/codec.py (class renamed FishCodec, loader trimmed), ogg/audio.py, ogg/editor.py, ogg/qwen3_asr/*, web/editor_core.js and | |
| web/ogg_editor.js (renamed from drama3_editor.js), and the structure of the voice, codes, decode, editor and ASR nodes in nodes.py. | |
| Node ids, value types, the editor route and the web extension were renamed (Ogg*, OGG_*, /ogg/editor/plan) so that this pack and | |
| ComfyUI-Drama3 can be installed side by side. The two example workflows follow its two. | |
| Its own notices apply to what it took from others: the codec is a port of Fish Audio's S2-Pro codec (Fish Audio Community | |
| License, LICENSE-FISH-AUDIO); the Qwen3-ASR code follows Qwen's qwen-asr (Apache-2.0). | |
| 3. Dialogue node: the idea of named roles each bound to a reference voice follows the description in the README of | |
| https://github.com/T8mars/Comfyui-Drama-3 (GPL-3.0). Its source could not be read when OggDialogue was written; no code of it is used. | |
| The node is built on OGG TTS's own dialogue code (voice_context, dialogue_text). | |
| 4. Fish Audio materials: Fish Audio Materials are licensed under the Fish Audio Community License, Version 1.0. | |
| Copyright © 39 AI, INC. All Rights Reserved. Built with Fish Audio. Not affiliated with, sponsored or endorsed by it. | |
| 5. OmniVoice (k2-fsa): its config.json and tokenizer files are read from the omnivoice folder, under its own license. | |
| Weights are not part of this pack: ogg-tts model.safetensors, the codec (drama3_codec.safetensors) and Qwen3-ASR are downloaded by the | |
| user, each under its own license. Use of the codec and of OGG TTS output is subject to the Fish Audio Community License, including its | |
| restriction on cloning the voice of a real person without consent. | |