336 lines
17 KiB
Python
336 lines
17 KiB
Python
"""
|
||
Runware Audio Inference Settings Node
|
||
Provides lyrics, guidanceType, languageBoost, turbo, temperature, textNormalization,
|
||
bpm, keyScale, timeSignature, vocalLanguage, coverConditioningScale, repaintingStart,
|
||
repaintingEnd, xVectorOnly, maxNewTokens, transcript, and more for Runware Audio Inference.
|
||
"""
|
||
|
||
from typing import Dict, Any
|
||
|
||
_VOCAL_LANGUAGE_CODES = [
|
||
"ar", "az", "bg", "bn", "ca", "cs", "da", "de", "el", "en", "es", "fa", "fi", "fr",
|
||
"he", "hi", "hr", "ht", "hu", "id", "is", "it", "ja", "ko", "la", "lt", "ms", "ne",
|
||
"nl", "no", "pa", "pl", "pt", "ro", "ru", "sa", "sk", "sr", "sv", "sw", "ta", "te",
|
||
"th", "tl", "tr", "uk", "ur", "vi", "yue", "zh", "unknown",
|
||
]
|
||
|
||
|
||
class RunwareAudioSettings:
|
||
"""Runware Audio Inference Settings Node."""
|
||
|
||
@classmethod
|
||
def INPUT_TYPES(cls):
|
||
return {
|
||
"required": {},
|
||
"optional": {
|
||
"useTemperature": ("BOOLEAN", {
|
||
"tooltip": "Enable to include temperature in audio generation settings",
|
||
"default": False,
|
||
}),
|
||
"temperature": ("FLOAT", {
|
||
"tooltip": "Sampling temperature for text token generation. Lower values produce more coherent, predictable speech content. Higher values increase variety but risk artifacts. Set 0 for greedy decoding. Only used when 'Use Temperature' is enabled.",
|
||
"default": 0.6,
|
||
"min": 0.0,
|
||
"max": 5.0,
|
||
"step": 0.1,
|
||
}),
|
||
"useAudioTemperature": ("BOOLEAN", {
|
||
"tooltip": "Enable to include audioTemperature in audio generation settings",
|
||
"default": False,
|
||
}),
|
||
"audioTemperature": ("FLOAT", {
|
||
"tooltip": "Sampling temperature for audio token generation. Lower values produce cleaner, more stable audio. Higher values add expressiveness but may introduce noise. Set 0 for greedy decoding. Only used when 'Use Audio Temperature' is enabled.",
|
||
"default": 0.8,
|
||
"min": 0.0,
|
||
"max": 5.0,
|
||
"step": 0.1,
|
||
}),
|
||
"useTopK": ("BOOLEAN", {
|
||
"tooltip": "Enable to include topK token sampling in audio generation settings",
|
||
"default": False,
|
||
}),
|
||
"topK": ("INT", {
|
||
"tooltip": "Top-K candidates for both text and audio token sampling. Limits sampling to the K most likely next tokens at each step. Only used when 'Use Top K' is enabled.",
|
||
"default": 50,
|
||
"min": 1,
|
||
"max": 500,
|
||
"step": 1,
|
||
}),
|
||
"useIncludePrefix": ("BOOLEAN", {
|
||
"tooltip": "Enable to include includePrefix in audio generation settings",
|
||
"default": False,
|
||
}),
|
||
"includePrefix": ("BOOLEAN", {
|
||
"tooltip": "When voice cloning, keep prefix audio in output (true) or trim to new speech (false). Only relevant when compatible prefix/reference audio is provided. Only used when 'Use Include Prefix' is enabled.",
|
||
"default": False,
|
||
"label_on": "true",
|
||
"label_off": "false",
|
||
}),
|
||
"useLyrics": ("BOOLEAN", {
|
||
"tooltip": "Enable to include lyrics in audio generation settings",
|
||
"default": False,
|
||
}),
|
||
"lyrics": ("STRING", {
|
||
"multiline": True,
|
||
"default": "",
|
||
"placeholder": "[verse] ... [chorus] ... (format like genius.com)",
|
||
"tooltip": "Lyrics for the audio generation. Format like lyrics from genius.com with [verse], [chorus], [refrain], etc.",
|
||
}),
|
||
"useGuidanceType": ("BOOLEAN", {
|
||
"tooltip": "Enable to include guidanceType (controls how guidance value is used)",
|
||
"default": False,
|
||
}),
|
||
"guidanceType": (["apg", "cfg"], {
|
||
"tooltip": "Controls how the guidance value is used. See API GuidanceType: apg | cfg.",
|
||
"default": "apg",
|
||
}),
|
||
"useLanguageBoost": ("BOOLEAN", {
|
||
"tooltip": "Enable to include languageBoost (language detection setting) in audio generation settings",
|
||
"default": False,
|
||
}),
|
||
"languageBoost": ([
|
||
"auto", "zh", "yue", "en", "ar", "ru", "es", "fr", "pt", "de", "tr", "nl", "uk",
|
||
"vi", "id", "ja", "it", "ko", "th", "pl", "ro", "el", "cs", "fi", "hi", "bg", "da",
|
||
"he", "ms", "fa", "sk", "sv", "hr", "fil", "hu", "no", "sl", "ca", "nn", "ta", "af",
|
||
], {
|
||
"tooltip": "Language detection setting. Codes: zh, yue, en, ar, ru, es, fr, pt, de, tr, nl, uk, vi, id, ja, it, ko, th, pl, ro, el, cs, fi, hi, bg, da, he, ms, fa, sk, sv, hr, fil, hu, no, sl, ca, nn, ta, af, auto",
|
||
"default": "auto",
|
||
}),
|
||
"useTurbo": ("BOOLEAN", {
|
||
"tooltip": "Enable to include turbo (select turbo model or not) in audio generation settings",
|
||
"default": False,
|
||
}),
|
||
"turbo": ("BOOLEAN", {
|
||
"tooltip": "When enabled, use turbo model. Only used when 'Use Turbo' is enabled.",
|
||
"default": False,
|
||
"label_on": "true",
|
||
"label_off": "false",
|
||
}),
|
||
"useTextNormalization": ("BOOLEAN", {
|
||
"tooltip": "Enable to include textNormalization (normalize text before synthesis) in audio generation settings",
|
||
"default": False,
|
||
}),
|
||
"textNormalization": ("BOOLEAN", {
|
||
"tooltip": "Normalize text (e.g. numbers, abbreviations) before synthesis to improve pronunciation. Only used when 'Use Text Normalization' is enabled.",
|
||
"default": True,
|
||
"label_on": "true",
|
||
"label_off": "false",
|
||
}),
|
||
"useBpm": ("BOOLEAN", {
|
||
"tooltip": "Enable to include bpm (tempo) in audio generation settings. If disabled, the model decides automatically.",
|
||
"default": False,
|
||
}),
|
||
"bpm": ("INT", {
|
||
"tooltip": "Beats per minute (30–300). If not set via API, the model decides. Common: 60–80 ballad, 90–110 pop, 120–130 house, 140–170 DnB. Only used when 'Use Bpm' is enabled.",
|
||
"default": 120,
|
||
"min": 30,
|
||
"max": 300,
|
||
"step": 1,
|
||
}),
|
||
"useKeyScale": ("BOOLEAN", {
|
||
"tooltip": "Enable to include keyScale (musical key and scale). Empty value lets the model decide.",
|
||
"default": False,
|
||
}),
|
||
"keyScale": ("STRING", {
|
||
"default": "",
|
||
"tooltip": "Musical key and scale. Empty string lets the model decide. Format: '{Note}{Accidental} {Mode}' (Note A–G, Accidental '' | '#' | 'b' | '♯' | '♭', Mode 'major' | 'minor'). Examples: 'C major', 'F# minor', 'B♭ major'.",
|
||
}),
|
||
"useTimeSignature": ("BOOLEAN", {
|
||
"tooltip": "Enable to include timeSignature (beats per measure). If disabled, the model decides automatically.",
|
||
"default": False,
|
||
}),
|
||
"timeSignature": ("INT", {
|
||
"default": 4,
|
||
"min": 1,
|
||
"max": 32,
|
||
"step": 1,
|
||
"tooltip": "Time signature as integer beats per bar (e.g. 2, 3, 4, 6). Only used when 'Use Time Signature' is enabled.",
|
||
}),
|
||
"useVocalLanguage": ("BOOLEAN", {
|
||
"tooltip": "Enable to include vocalLanguage (language code) for vocal generation. See dropdown for available codes.",
|
||
"default": False,
|
||
}),
|
||
"vocalLanguage": (_VOCAL_LANGUAGE_CODES, {
|
||
"tooltip": "Language code for vocals. Supports values from the dropdown list (including 'yue', 'zh', and 'unknown'). Use 'unknown' for instrumental or automatic detection. Only used when 'Use Vocal Language' is enabled.",
|
||
"default": "en",
|
||
}),
|
||
"useCoverConditioningScale": ("BOOLEAN", {
|
||
"tooltip": "Enable to include coverConditioningScale (source-audio conditioning fraction for cover tasks). Only effective when input audio is provided.",
|
||
"default": False,
|
||
}),
|
||
"coverConditioningScale": ("FLOAT", {
|
||
"tooltip": "Fraction of denoising steps using source-audio conditioning (cover task). 1.0 = all steps; 0.5 = first half source then text-only. Only used when 'Use Cover Conditioning Scale' is enabled.",
|
||
"default": 1.0,
|
||
"min": 0.0,
|
||
"max": 1.0,
|
||
"step": 0.01,
|
||
}),
|
||
"useRepaintingStart": ("BOOLEAN", {
|
||
"tooltip": "Enable to include repaintingStart (seconds). Requires input audio. Sets repaint region with repaintingEnd; negative values prepend (outpaint before start).",
|
||
"default": False,
|
||
}),
|
||
"repaintingStart": ("FLOAT", {
|
||
"tooltip": "Start time in seconds for repaint/inpaint/outpaint region. Only used when 'Use Repainting Start' is enabled.",
|
||
"default": 0.0,
|
||
"step": 0.01,
|
||
}),
|
||
"useRepaintingEnd": ("BOOLEAN", {
|
||
"tooltip": "Enable to include repaintingEnd (seconds). Requires input audio. Values beyond source duration append (outpaint).",
|
||
"default": False,
|
||
}),
|
||
"repaintingEnd": ("FLOAT", {
|
||
"tooltip": "End time in seconds for repaint region. Only used when 'Use Repainting End' is enabled.",
|
||
"default": 0.0,
|
||
"step": 0.01,
|
||
}),
|
||
"useXVectorOnly": ("BOOLEAN", {
|
||
"tooltip": "Enable to include xVectorOnly in audio generation settings.",
|
||
"default": False,
|
||
}),
|
||
"xVectorOnly": ("BOOLEAN", {
|
||
"tooltip": "true = speaker embedding only (no transcript needed, lower similarity). false = ICL mode (needs transcript, higher quality). Only used when 'Use X Vector Only' is enabled.",
|
||
"default": False,
|
||
"label_on": "true",
|
||
"label_off": "false",
|
||
}),
|
||
"useMaxNewTokens": ("BOOLEAN", {
|
||
"tooltip": "Enable to include maxNewTokens (audio output token cap) in audio generation settings.",
|
||
"default": False,
|
||
}),
|
||
"maxNewTokens": ("INT", {
|
||
"tooltip": "Audio output token cap. Higher = longer audio, but risk of generation hangs. Applies to all variants. Only used when 'Use Max New Tokens' is enabled.",
|
||
"default": 4096,
|
||
"min": 1,
|
||
"max": 262144,
|
||
"step": 1,
|
||
}),
|
||
"useTranscript": ("BOOLEAN", {
|
||
"tooltip": "Enable to include transcript (reference audio transcript) in audio generation settings.",
|
||
"default": False,
|
||
}),
|
||
"transcript": ("STRING", {
|
||
"multiline": True,
|
||
"default": "",
|
||
"tooltip": "Transcript of the reference audio. Required for ICL mode; omit or leave empty for x-vector-only mode. Base variant only. Only used when 'Use Transcript' is enabled.",
|
||
}),
|
||
}
|
||
}
|
||
|
||
RETURN_TYPES = ("RUNWAREAUDIOSETTINGS",)
|
||
RETURN_NAMES = ("settings",)
|
||
FUNCTION = "createSettings"
|
||
CATEGORY = "Runware/Audio"
|
||
DESCRIPTION = (
|
||
"Configure audio generation settings (lyrics, guidanceType, languageBoost, turbo, temperature, textNormalization, "
|
||
"bpm, keyScale, timeSignature, vocalLanguage, coverConditioningScale, repaintingStart, repaintingEnd, "
|
||
"xVectorOnly, maxNewTokens, transcript, etc.) for Runware Audio Inference. Connect to Runware Audio Inference node."
|
||
)
|
||
|
||
def createSettings(self, **kwargs) -> tuple[Dict[str, Any]]:
|
||
"""Create audio settings dict for API"""
|
||
use_temperature = kwargs.get("useTemperature", False)
|
||
temperature = kwargs.get("temperature", 0.6)
|
||
use_audio_temperature = kwargs.get("useAudioTemperature", False)
|
||
audio_temperature = kwargs.get("audioTemperature", 0.8)
|
||
use_top_k = kwargs.get("useTopK", False)
|
||
top_k = kwargs.get("topK", 50)
|
||
use_include_prefix = kwargs.get("useIncludePrefix", False)
|
||
include_prefix = kwargs.get("includePrefix", False)
|
||
use_lyrics = kwargs.get("useLyrics", False)
|
||
lyrics = kwargs.get("lyrics", "")
|
||
use_guidance_type = kwargs.get("useGuidanceType", False)
|
||
guidance_type = kwargs.get("guidanceType", "apg")
|
||
use_language_boost = kwargs.get("useLanguageBoost", False)
|
||
language_boost = kwargs.get("languageBoost", "auto")
|
||
use_turbo = kwargs.get("useTurbo", False)
|
||
turbo = kwargs.get("turbo", False)
|
||
use_text_normalization = kwargs.get("useTextNormalization", False)
|
||
text_normalization = kwargs.get("textNormalization", True)
|
||
use_bpm = kwargs.get("useBpm", False)
|
||
bpm = kwargs.get("bpm", 120)
|
||
use_key_scale = kwargs.get("useKeyScale", False)
|
||
key_scale = kwargs.get("keyScale", "")
|
||
use_time_signature = kwargs.get("useTimeSignature", False)
|
||
time_signature = kwargs.get("timeSignature", 4)
|
||
use_vocal_language = kwargs.get("useVocalLanguage", False)
|
||
vocal_language = kwargs.get("vocalLanguage", "en")
|
||
use_cover_conditioning_scale = kwargs.get("useCoverConditioningScale", False)
|
||
cover_conditioning_scale = kwargs.get("coverConditioningScale", 1.0)
|
||
use_repainting_start = kwargs.get("useRepaintingStart", False)
|
||
repainting_start = kwargs.get("repaintingStart", 0.0)
|
||
use_repainting_end = kwargs.get("useRepaintingEnd", False)
|
||
repainting_end = kwargs.get("repaintingEnd", 0.0)
|
||
use_x_vector_only = kwargs.get("useXVectorOnly", False)
|
||
x_vector_only = kwargs.get("xVectorOnly", False)
|
||
use_max_new_tokens = kwargs.get("useMaxNewTokens", False)
|
||
max_new_tokens = kwargs.get("maxNewTokens", 4096)
|
||
use_transcript = kwargs.get("useTranscript", False)
|
||
transcript = kwargs.get("transcript", "") or ""
|
||
|
||
settings: Dict[str, Any] = {}
|
||
|
||
if use_temperature:
|
||
settings["temperature"] = float(temperature)
|
||
if use_audio_temperature:
|
||
settings["audioTemperature"] = float(audio_temperature)
|
||
if use_top_k:
|
||
settings["topK"] = int(top_k)
|
||
if use_include_prefix:
|
||
settings["includePrefix"] = bool(include_prefix)
|
||
|
||
if use_lyrics and lyrics and lyrics.strip():
|
||
settings["lyrics"] = lyrics.strip()
|
||
|
||
if use_guidance_type:
|
||
settings["guidanceType"] = guidance_type
|
||
|
||
if use_language_boost:
|
||
settings["languageBoost"] = language_boost
|
||
|
||
if use_turbo:
|
||
settings["turbo"] = bool(turbo) # API requires true/false
|
||
|
||
if use_text_normalization:
|
||
settings["textNormalization"] = bool(text_normalization)
|
||
|
||
if use_bpm:
|
||
settings["bpm"] = int(bpm)
|
||
|
||
if use_key_scale and str(key_scale).strip():
|
||
settings["keyScale"] = str(key_scale).strip()
|
||
|
||
if use_time_signature:
|
||
settings["timeSignature"] = int(time_signature)
|
||
|
||
if use_vocal_language:
|
||
settings["vocalLanguage"] = vocal_language
|
||
|
||
if use_cover_conditioning_scale:
|
||
settings["coverConditioningScale"] = float(cover_conditioning_scale)
|
||
|
||
if use_repainting_start:
|
||
settings["repaintingStart"] = float(repainting_start)
|
||
|
||
if use_repainting_end:
|
||
settings["repaintingEnd"] = float(repainting_end)
|
||
|
||
if use_x_vector_only:
|
||
settings["xVectorOnly"] = bool(x_vector_only)
|
||
|
||
if use_max_new_tokens:
|
||
settings["maxNewTokens"] = int(max_new_tokens)
|
||
|
||
if use_transcript and transcript.strip():
|
||
settings["transcript"] = transcript.strip()
|
||
|
||
return (settings,)
|
||
|
||
|
||
NODE_CLASS_MAPPINGS = {
|
||
"RunwareAudioSettings": RunwareAudioSettings,
|
||
}
|
||
|
||
NODE_DISPLAY_NAME_MAPPINGS = {
|
||
"RunwareAudioSettings": "Runware Audio Inference Settings",
|
||
}
|