LipsyncTTS: only add TTS cost when TTS provider is 11 labs

GooeyAI · Oct 4, 2023 · c0a239d · c0a239d
1 parent 087e330
commit c0a239d
Show file tree

Hide file tree

Showing 2 changed files with 19 additions and 12 deletions.
diff --git a/recipes/LipsyncTTS.py b/recipes/LipsyncTTS.py
@@ -1,3 +1,4 @@
+import textwrap
 import typing
 
 from pydantic import BaseModel
@@ -6,7 +7,7 @@
 from bots.models import Workflow
 from recipes.DeforumSD import safety_checker
 from recipes.Lipsync import LipsyncPage
-from recipes.TextToSpeech import TextToSpeechPage
+from recipes.TextToSpeech import TextToSpeechPage, TextToSpeechProviders
 from daras_ai_v2.loom_video_widget import youtube_video
 
 DEFAULT_LIPSYNC_TTS_META_IMG = "https://storage.googleapis.com/dara-c1b52.appspot.com/daras_ai/media/assets/lipsync_meta_img.gif"
@@ -150,17 +151,23 @@ def render_output(self):
         self.render_example(st.session_state)
 
     def get_raw_price(self, state: dict):
-        return LipsyncPage.get_raw_price(self, state) + \
-                TextToSpeechPage.get_raw_price(self, state)
+        # _get_tts_provider comes from TextToSpeechPage
+        if self._get_tts_provider(state) == TextToSpeechProviders.ELEVEN_LABS:
+            return LipsyncPage.get_raw_price(self, state) + \
+                    TextToSpeechPage.get_raw_price(self, state)
+        else:
+            return LipsyncPage.get_raw_price(self, state)
 
     def additional_notes(self):
-        notes = f"""
-*Cost = Lipsync Cost + TTS Cost*
-"""
-        if lipsync_notes := LipsyncPage.additional_notes(self):
-            notes += "\n" + f"*Lipsync* {lipsync_notes}"
+        lipsync_notes = LipsyncPage.additional_notes(self)
         if tts_notes := TextToSpeechPage.additional_notes(self):
-            notes += "\n" + f"*TTS* {tts_notes}"
+            notes = textwrap.dedent(f"""\
+                - *Lipsync* {lipsync_notes.strip()}
+                - *TTS* {tts_notes.strip()}
+            """)
+        else:
+            notes = lipsync_notes
+
         return notes
 
     def render_usage_guide(self):

diff --git a/recipes/TextToSpeech.py b/recipes/TextToSpeech.py
@@ -126,8 +126,8 @@ def render_output(self):
 
     def _get_eleven_labs_price(self, state: dict):
         text = state.get("text_prompt", "")
-        # 4 credits for 10 words ~ 50 chars
-        return (len(text) / 50) * 4
+        # 0.079 credits / character ~ 4 credits / 10 words
+        return len(text) * 0.079
 
     def _get_tts_provider(self, state: dict):
         tts_provider = state.get(
@@ -140,7 +140,7 @@ def additional_notes(self):
         tts_provider = st.session_state.get("tts_provider")
         if tts_provider == TextToSpeechProviders.ELEVEN_LABS.name:
             return """
-    *Eleven Labs cost ≈ 4 credits for every 50 characters ≈ 0.4 credits per word*
+    *Eleven Labs cost ≈ 4 credits per 10 words*
             """
         else:
             return ""