Learnings 11labs prompting v3

USE OF SINGLE INVERTED COMMAS TO MAKE AN EMPHASIS
https://studio-internal.scenarix.ai/assetGen?tool=audioGen\&v2=true\&sharedJobView=2e129126-a7e9-4700-afa7-bab2199e640f

USE OF BOLD CHARACTERS TO MAKE A STRONG EMPHASIS
https://studio-internal.scenarix.ai/assetGen?tool=audioGen\&v2=true\&sharedJobView=2be52f09-80cf-45d7-a23b-5bc4b3c7d10c

USE OF NO SPACES BETWEEN WORDS FOR VARIATION IN SPEED
https://studio-internal.scenarix.ai/assetGen?tool=audioGen\&v2=true\&sharedJobView=26442981-8f3c-409c-bdf9-2a82c3a9c5d7 (NOT JOINED)
https://studio-internal.scenarix.ai/assetGen?tool=audioGen\&v2=true\&sharedJobView=f306de6f-3381-4e66-9718-82604638d723 (JOINED)

USE OF ALL BOLD CHARACTERS FOR THROW OF VOICE
https://studio-internal.scenarix.ai/assetGen?tool=audioGen\&v2=true\&sharedJobView=42049fed-619e-4d00-a9c8-edf8be728645

USE OF PARENTHETICALS IN BETWEEN DIALOGUES TO CREATE VARIATION OF EMOTION, PITCH AND SPEED IN DELIVERY
https://studio-internal.scenarix.ai/assetGen?tool=audioGen\&v2=true\&sharedJobView=f239d78f-b0dd-464a-82f9-014122c4098a

Effective Prompt Structure:

  • Each brackets affects the following text, until the next bracket
  • Example: Play [dark][demon]Hi guys, and 'welcome' to the Dark Lord's 'TTS tutorial' [long and evil laugh][laugh fading]...[clears throat][loud]'My' name is Kyle, and I'll guide you through dialogue creation... [menacing][louder]read this through and play the audio... [shout][angry][screams] but fail to do that andYOU will se me in your DREAMS!
  • Separate square brackets work better than comma-separated tags
  • Example: [anxious][trembling voice][panting breathing]Who are you.. and what are you doing in 'my' room? VS[anxious,trembling voice,panting breathing]Who are you.. and what are you doing in 'my' room?
  • Complex emotional transitions possible within single generation
  • Example: Play Can I please come with you? [pleading] Please, please, please [desperate] Pleaaaaaase.
  • Pauses e.g. [pause][long pause] work better than typographic additions

Advanced Emotion Control Techniques:

  • Typography impacts:
  • Emphasis: use single quotes around word
    • Play and what are you doing in ‘my’ room?
  • Strong emphasis: capitalize word
    • Play and what are you doing in ‘MY’ room?
  • Strongest emphasis and higher volume: use question mark (1 is enough) after the target word
    • Play and what are you doing in ‘MY’! Room?
  • “...” inserts pauses in between
    • Play [terrified] Kyle! ... Kyle, what’s happening?
    • Combining “…” in series does not make for longer pauses - to be reviewed
  • Line break + … for longer pauses
  • Sound effects: Whimpers, sighs before each part of the dialogue
  • Example:Play [sigh][confused]Alone!? [sigh]..[matter of fact][fast] If you wanted a night alone, [fast][irritated] why invite all our friends?
  • Tone progression and climax: Scared → loud → shouting in single generation
  • Example:Play [panting][scared][calling out] Kyle..?[panting][questioning][loud]Can you hear me? [shout] [panting] Kyle! [panting]
  • Repeating letter multiple times in the tail:
  • Can be useful for emulating shouting
    (example: https://studio-internal.scenarix.ai/assetGen?tool=audioGen&sharedJobView=eedd9186-c1a9-4402-830f-f75766a2db7e)
  • Attach words to get different speed
  • [WORRIED] [SCARED] RIGHT HERE! [SCARED] [TREMBLING] ThisiswhereKyle ATTACKED ME