Unsloth Model Card

README.md

@@ -9,47 +9,7 @@ license: apache-2.0
 language:
 - en
 ---
-Supports a single voice id 0. Features:
-- Tagging. Laugh and sigh work best.
-  -   \[sigh\]: 1,303
-  -   \[laugh\]: 630
-  -   \[sighs\]: 558
-  -   \[laughs\]: 254
-  -   \[gasp\]: 136
-  -   \[gasps\]: 56
-  -   \[sniff\]: 108
-  -   \[um\]: 93
-  -   \[scoff\]: 91
-  -   \[mhm\]: 51
-  -   \[uh\]: 48
-  -   \[sniffs\]: 46
-  -   \[applause\]: 46
-  -   \[scoffs\]: 45
-  -   \[cough\]: 45
-  -   \[woo\]: 43
-  -   \[hum\]: 39
-  -   \[sting\]: 26
-  -   \[breath\]: 26
-  -   \[uh-huh\]: 23
-  -   \[coughs\]: 23
-- Reading out numbers and emails. Ensure to delimit numbers (should be able to handle any delimiter)
-- Single or few word responses
-- Very short (or single word) questions
-- Speaker context for consistency
-- Phonetic pronounciation (no formal notation though)
-- For now, because it is single speaker, does NOT need an initial grounding example, but context building is recommended.
 
-Known issues:
-- do not put a tag after numbers
-- cannot do times with colons (convert 10:00 am-> 10 am and convert 10:10 am -> 10 10 am) (this will be fixed with another version soon)
-
-Upcoming features:
-- Robust time reading (n + 1)
-- Optional numerical tag for how natural it should sound. ie: "How are you?": (n + 2)
-  - To make it sound human do: "\[10\] How are you?"
-  - Or to make it robotic do: "\[1\] How are you?"
-  - If not set, defaults to ~9 or 10.
-- Data cleanup of garbled samples. (n + 3)
 # Uploaded finetuned  model
 
 - **Developed by:** Cyleux