jhddlmZmZmZmZmZmZeddededgZeded<edd ed zd ed zd gZedded<eddedzdddgZ eddde d<eddedzdedzdgZ ede d<eddeded zd!gZ ede d<d"d#edd$Z d%Z d&d'eddd$Zd(S))) LYRICS_PROMPTLYRICS_PROMPT_SHORT SPEECH_PROMPTSPEECH_PROMPT_SHORT help_text specialitieszChatterbox SpeechaSpeak the text in the selected **Language**, optionally using **Voice to Replicate** as a speaker reference. Use clean single-speaker audio. **Emotion Exaggeration** increases expressiveness; **Pace** changes delivery. Begin at their neutral defaults and change one control at a time. This is the multilingual Chatterbox workflow; do not assume special tags advertised for other Chatterbox checkpoints are accepted.zprompt = spoken words; model_mode = language; optional audio_guide = voice sample. Exaggeration/Pace adjust delivery. Use this multilingual checkpoint's controls, not Turbo-specific tags.z)https://github.com/resemble-ai/chatterboxz voice cloningrzKugelAudio Speech and DialogueaGenerate speech from text, optionally cloning one to three reference voices. Choose **Text Only** or the matching **Voice Cloning** mode. References must be clean single-speaker recordings. For dialogue, number speakers in reference order. **Max Duration** caps the assembled result. Use blank lines to split long passages; **Auto Split** helps avoid speech accelerating in long segments, and **Pause Between Sentences** adds spacing.z ### Dialogue ```text Speaker 1: Shall we begin? Speaker 2: Yes, I have everything ready. ``` Use `Speaker 3:` only when a third reference voice is supplied. Write in the desired language; this family focuses on European-language speech.zText only, or audio_prompt_type A/AB/ABD for 1/2/3 reference voices. Speaker N labels follow reference order. Max Duration caps speech; blank lines and Auto Split divide passages; pauses add spacing.z\ For dialogue use `Speaker 1: ...` / `Speaker 2: ...` / `Speaker 3: ...` in reference order.z-https://github.com/Kugelaudio/kugelaudio-openzspeech dialoguezOmniVoice SpeechakGenerate speech from a voice description or clone one to three reference voices. Select the matching voice mode and language. **Max Duration** limits the output (0 selects automatic duration). Keep a reference short, clear and from one speaker. In cloning mode the secondary text field holds the reference transcript; in design mode it holds supported voice tags.aC ### Voice and Dialogue In **Voice Design**, put supported comma-separated tags in **Voice Instruction**, for example `female, young adult, british accent`. In **Voice Cloning**, put the exact reference transcript in that field, or leave it blank for automatic transcription. Dialogue uses `Speaker 1:` / `Speaker 2:` / `Speaker 3:` lines matching the supplied references. Supported non-verbal tags such as `[laughter]` and `[sigh]` belong in the spoken text; the full list appears in the model input help. **Example:** `[sigh] We finally made it. [laughter] What a journey!`uVoice design: supported comma-separated tags in alt_prompt. Cloning: A/AB/ABD select 1/2/3 audio references; alt_prompt is the exact reference transcript(s), blank for transcription. Prefer 3–10s clean samples in target language. Max Duration 0 = auto.aprompt contains speech; dialogue uses Speaker N labels matching audio references. [laughter]/[sigh] and the documented non-verbal tags are supported. Do not invent emotion tags. Voice-design tags go in alt_prompt; cloning transcripts describe the reference, not target speech.z#https://github.com/k2-fsa/OmniVoicez voice designzHeartMuLa SongsaCombine **Lyrics** with comma-separated **Keywords / Tags** for genre, mood, instruments and vocal character. **Maximum Song Duration** is a cap: the song may finish earlier, or be cut off if it needs more time. Begin with a short verse and chorus, then extend the lyrics and duration. This entry generates from text; it does not expose the reference-audio or MIDI workflows of separate HeartMuLa projects.z **Keywords / Tags example:** `acoustic folk,warm,female vocals,guitar,soft drums`. For an instrumental, use an `[Instrumental]` section and instrumental style tags.zprompt = sectioned lyrics; alt_prompt = comma-separated sound/style tags. Maximum Song Duration caps output. No audio/MIDI reference input for this entry.zx alt_prompt example: acoustic folk,warm,female vocals,guitar. For instrumental use [Instrumental] and instrumental tags.z%https://github.com/HeartMuLa/heartlibzmusic generationzACE-Step 1.0 MusicaKCombine **Lyrics** and **Genres / Tags** to create a song, or use `[Instrumental]` lyrics for music without singing. Duration selects the target length. **Remix Audio** uses Source Audio, the original lyrics and **Prompt Audio Strength** to guide a variation. Keep this 1.0 workflow separate from ACE-Step 1.5's cover/timbre modes.zACE-Step 1.0: prompt lyrics, alt_prompt genre/style tags, target duration. A remixes audio_guide with original lyrics and audio strength. No separate 1.5 timbre mode.zK Instrumental: [Instrumental] lyrics and an instrumental style description.z$https://github.com/ace-step/ACE-Stepa### MiniMax Music 3 Lyrics in prompt, music description in alt_prompt. Maximum Song Duration is a cap (up to 300s); generation can end sooner. Text-conditioned songs, no reference-audio control here. Tags and tempo/key requests are generative guidance, not exact guarantees.a### Lyrics and Caption Use actual lyrics with section tags on their own lines: [Verse], [Chorus], [Bridge], [Outro]. Keep genre, mood, voices, instruments and development in alt_prompt. Natural language works; a richer caption can use Global Metadata, Vocal Details and Arrangement. Keep lyric words in prompt. Example caption: warm acoustic pop, intimate vocals, fingerpicked guitar, soft percussion building into the chorus.) deepy_infosdeepy_prompt_infosra### Speaker References and Dialogue Supply a clean **Speaker Reference Voice** and write the words to speak. **Voice Cloning** uses that recording's voice and emotion. **Voice + Emotion** uses a second recording for emotion while retaining the first voice. In **Dialogue**, provide two or three speaker recordings and label the text `Speaker 1:`, `Speaker 2:` and `Speaker 3:` in reference order. **Default Emotion Instruction** and inline emotion cues control delivery separately from the words. **Max Duration** caps the assembled output.a4### IndexTTS 2.5 Required speaker audio. A clones it; AB adds emotion audio; AB2/ABD2 use 2/3 speaker references. prompt = speech, alt_prompt = default emotion; inline cues override emotion per segment. Select Language: EN/ZH/ZHEN/JA/ES/AR. Speech Speed 1 is normal, higher faster. Max duration caps output.un### Speech, Emotion and Pronunciation Write exact speech; dialogue uses Speaker N: labels in sample order. Inline [happy] / [calm] cues persist until replaced; alt_prompt supplies default emotion. Pronunciation: Chinese <行|XING2>, English , Japanese <上手|じょうず>. Select matching language. Keep unrelated directions out of speech.N)models.help_sharedrrrrrrCHATTERBOX_HELP KUGEL_HELPOMNIVOICE_HELPHEARTMULA_HELP ACE1_HELPMINIMAX_MUSIC_HELPINDEX_VOICE_INPUTS INDEX25_HELP7C:\Users\johnn\OneDrive\Documents\AI\models\TTS\help.pyrs~~~~~~~~~~~~~~~~)/2SUbda c v y d x e f f ".,"?"? Y7:np}@{ p{ } FH[^|H|n~opp )\/;LMM >-0]_lo|_|~~@VY~X@@!-o~O`!a!a~,/GIVYC IC E a c v y sc sv]u^__!-.@!A!A~ I*-z|IKsuH K X uX [ A Z B C C (L);<< .kJ L!344 SLM L2CDD r