Upload 2 files

README.md

@@ -1,6 +1,6 @@
 ---
 title: Misaki G2P
-emoji: 🐨
+emoji: ⚡
 colorFrom: yellow
 colorTo: gray
 sdk: gradio
@@ -11,4 +11,4 @@ license: apache-2.0
 short_description: G2P
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,165 @@
+from gradio_client import Client
+import gradio as gr
+import json
+import os
+import pprint
+import spaces
+
+@spaces.GPU
+def greet(n):
+    return f"Hello {zero + n} Tensor"
+
+client = Client('hexgrad/misaki-src', hf_token=os.environ['TOKEN'])
+def predict(text, use_spacy_transformer, british):
+    phonemes, trace, elapsed_cpu_time = client.predict(text=text, trf=use_spacy_transformer, british=british, api_name='/predict')
+    trace = pprint.pformat(json.loads(trace))
+    return phonemes, trace, elapsed_cpu_time
+
+with gr.Blocks() as app:
+    gr.Markdown('''
+Misaki is an experimental G2P engine designed to power future versions of Kokoro models.
+
+This English-only preview is primarily intended for researchers and linguists. It may be deeply uninteresting to most people.
+''', container=True)
+    gr.Interface(fn=predict, inputs=[gr.Text(), gr.Checkbox(), gr.Checkbox()],
+                 outputs=[gr.Text(label='phonemes'), gr.Text(label='trace'), gr.Number(label='elapsed_cpu_time')])
+    gr.Markdown('''
+### Examples
+```md
+American: [Misaki](/misˈɑki/) is an experimental G2P engine designed to power future versions of [Kokoro](/kˈOkəɹO/) models.
+
+British: [Misaki](/misˈɑːki/) is an experimental G2P engine designed to power future versions of [Kokoro](/kˈQkəɹQ/) models.
+
+But I am the Chosen One. But I [am](+1) the Chosen [One](-1). But I [am](+2) the Chosen [One](-2).
+
+1002. [1002](#a#). [1002](#an#). [1002](#a&#). 2025. 2,025. $45.67 billion trillion.
+```
+''', container=True)
+    gr.Markdown('''
+### Token-Level Trace
+```py
+# 1. Text. Can be useful for aligning text to phonemes, e.g. highlighting text during audio playback.
+
+# 2. Tag. See a full list of tags from spaCy:
+# https://github.com/explosion/spaCy/blob/master/spacy/glossary.py
+
+# 3. Whitespace. Whether or not a token has trailing whitespace (string => bool for this demo).
+whitespace = True if whitespace else False
+
+# 4. Phonemes. For this demo, the question mark means UNK, the ninja emoji means empty string.
+phonemes = '❓' if phonemes is None else ('🥷' if phonemes == '' else phonemes)
+
+# 5. Rating. Star rating for the estimated quality of this token's phonemes.
+ratings = dict(
+user_override = '💎(5/5)',
+gold = '🏆(4/5)',
+silver = '🥈(3/5)',
+bronze = '🥉(2/5)',
+unk = '❓(UNK)',
+)
+```
+''', container=True)
+    gr.Markdown('''
+### Notes
+- For English, Misaki uses a gold dictionary with 80k words and a similarly sized silver dictionary.
+- There are separate dictionaries for American & British English.
+- Users can override the dictionary and/or individual tokens with custom pronunciations.
+- `espeak-ng` is used as the fallback for OOD words, and the token is rated "bronze" in this case.
+- Raw token objects are returned, with phonemes aligned at the per-token level.
+- UNKs are easy to detect when `token.phonemes is None`.
+- The entire implementation of Misaki (English) is <1000 lines of Python, excluding dictionary files.
+- POS disambiguation should be live, e.g. to wound someone vs wound up.
+- use_spacy_transformer should deliver more reliable POS tags.
+- Non-POS-based disambiguation, like graph axes vs throwing axes, is still a TODO.
+''', container=True)
+
+with gr.Blocks() as info:
+    gr.Markdown('''
+# Misaki English Phonemes
+
+For English, Misaki currently uses 49 total phonemes. Of these, 41 are shared by both Americans and Brits, 4 are American-only, and 4 are British-only.
+
+Disclaimer: Author is an ML researcher, not a linguist, and may have butchered or reappropriated the traditional meaning of some symbols. These symbols are intended as input tokens for neural networks to yield optimal performance.
+
+
+### 🤝 Shared (41)
+
+**Stress Marks (2)**
+- `ˈ`: Primary stress, visually looks similar to an apostrophe.
+- `ˌ`: Secondary stress.
+
+**IPA Consonants (22)**
+- `bdfhjklmnpstvwz`: 15 alpha consonants taken from IPA. They mostly sound as you'd expect, but `j` actually represents the "y" sound, like `yes => jˈɛs`.
+- `ɡ`: Hard "g" sound, like `get => ɡɛt`. Visually looks like the lowercase letter g, but its actually `U+0261`.
+- `ŋ`: The "ng" sound, like `sung => sˈʌŋ`.
+- `ɹ`: Upside-down r is just an "r" sound, like `red => ɹˈɛd`.
+- `ʃ`: The "sh" sound, like `shin => ʃˈɪn`.
+- `ʒ`: The "zh" sound, like `Asia => ˈAʒə`.
+- `ð`: Soft "th" sound, like `than => ðən`.
+- `θ`: Hard "th" sound, like `thin => θˈɪn`.
+
+**Consonant Clusters (2)**
+- `ʤ`: A "j" or "dg" sound, merges `dʒ`, like `jump => ʤˈʌmp` or `lunge => lˈʌnʤ`.
+- `ʧ`: The "ch" sound, merges `tʃ`, like `chump => ʧˈʌmp` or `lunch => lˈʌnʧ`.
+
+**IPA Vowels (10)**
+- `ə`: The schwa is a common, unstressed vowel sound, like `a 🍌 => ə 🍌`.
+- `i`: As in `easy => ˈizi`.
+- `u`: As in `flu => flˈu`.
+- `ɑ`: As in `spa => spˈɑ`.
+- `ɔ`: As in `all => ˈɔl`.
+- `ɛ`: As in `hair => hˈɛɹ` or `bed => bˈɛd`. Possibly dubious, because those vowel sounds do not sound similar to my ear.
+- `ɜ`: As in `her => hɜɹ`. Easy to confuse with `ɛ` above.
+- `ɪ`: As in `brick => bɹˈɪk`.
+- `ʊ`: As in `wood => wˈʊd`.
+- `ʌ`: As in `sun => sˈʌn`.
+
+**Dipthong Vowels (4)**
+- `A`: The "eh" vowel sound, like `hey => hˈA`. Expands to `eɪ` in IPA.
+- `I`: The "eye" vowel sound, like `high => hˈI`. Expands to `aɪ` in IPA.
+- `W`: The "ow" vowel sound, like `how => hˌW`. Expands to `aʊ` in IPA.
+- `Y`: The "oy" vowel sound, like `soy => sˈY`. Expands to `ɔɪ` in IPA.
+
+**Custom Vowel (1)**
+- `ᵊ`: Small schwa, muted version of `ə`, like `pixel => pˈɪksᵊl`. I made this one up, so I'm not entirely sure if it's correct.
+
+
+### 🇺🇸 American-only (4)
+
+**Vowels (3)**
+- `æ`: The vowel sound at the start of `ash => ˈæʃ`.
+- `O`: Capital letter representing the American "oh" vowel sound. Expands to `oʊ` in IPA.
+- `ᵻ`: A sound somewhere in between `ə` and `ɪ`, often used in certain -s suffixes like `boxes => bˈɑksᵻz`.
+
+**Consonant (1)**
+- `ɾ`: A sound somewhere in between `t` and `d`, like `butter => bˈʌɾəɹ`.
+
+
+### 🇬🇧 British-only (4)
+
+**Vowels (3)**
+- `a`: The vowel sound at the start of `ash => ˈaʃ`.
+- `Q`: Capital letter representing the British "oh" vowel sound. Expands to `əʊ` in IPA.
+- `ɒ`: The sound at the start of `on => ˌɒn`. Easy to confuse with `ɑ`, which is a shared phoneme.
+
+**Other (1)**
+- `ː`: Vowel extender, visually looks similar to a colon. Possibly dubious, because Americans extend vowels too, but the gold US dictionary somehow lacks these. Often used by the Brits instead of `ɹ`: Americans say `or => ɔɹ`, but Brits say `or => ɔː`.
+
+
+### ♻️ Misaki to espeak
+```py
+def to_espeak(ps):
+    # Optionally, you can add a tie character in between the 2 replacement characters.
+    ps = ps.replace('ʤ', 'dʒ').replace('ʧ', 'tʃ')
+    ps = ps.replace('A', 'eɪ').replace('I', 'aɪ').replace('Y', 'ɔɪ')
+    ps = ps.replace('O', 'oʊ').replace('Q', 'əʊ').replace('W', 'aʊ')
+    return ps.replace('ᵊ', 'ə')
+```
+''')
+
+demo = gr.TabbedInterface(
+    [app, info],
+    ['🔥 Misaki English Preview', 'ℹ️ Phonemes'],
+)
+
+demo.launch()