Michael/Socialflow
1
0 Fork 0
mirror of https://github.com/R0m1k3/Socialflow.git synced 2026-10-23 06:58:12 +02:00
Code Issues Packages Projects Releases Wiki Activity

feat(ffmpeg): Gemini TTS as primary engine, Edge TTS as fallback

Browse Source 
Default tts_engine changed from "edge" to "gemini".
Both /process-reel and /preview-tts now try Gemini first,
fall back to Edge TTS on failure.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
MichaelandClaude Opus 4.7 committed 2026-05-19 16:42:45 +02:00
1 parent 07fd43a5ed
commit c008b41eb7
1 file changed
+43 -20
Download Patch File Download Diff File Expand all files Collapse all files
ffmpeg-service/main.py
+43 -20
View File
@@ -159,7 +159,7 @@ class ReelRequest(BaseModel):
music_volume: float = 0.25 music_volume: float = 0.25
tts_enabled: bool = False tts_enabled: bool = False
tts_voice: str = "fr-FR-VivienneMultilingualNeural" tts_voice: str = "fr-FR-VivienneMultilingualNeural"
tts_engine: str = "edge" # "edge" or "gemini" tts_engine: str = "gemini" # "gemini" or "edge"
gemini_api_key: Optional[str] = None # Google Cloud API key for Gemini TTS gemini_api_key: Optional[str] = None # Google Cloud API key for Gemini TTS
draw_text: bool = True draw_text: bool = True
stabilize: bool = False # Stabilisation vidéo via vidstab stabilize: bool = False # Stabilisation vidéo via vidstab
@@ -924,18 +924,30 @@ async def process_reel(request: ReelRequest, x_api_key: str = Header(None)):
if tts_clean_text: if tts_clean_text:
print(f"🔊 Generating TTS audio to: {tts_audio_path}") print(f"🔊 Generating TTS audio to: {tts_audio_path}")
engine = request.tts_engine or "edge" engine = request.tts_engine or "gemini"
# Primary: Gemini TTS. Fallback: Edge TTS
if engine == "gemini" and request.gemini_api_key: if engine == "gemini" and request.gemini_api_key:
await generate_tts_gemini( try:
tts_clean_text, await generate_tts_gemini(
voice, tts_clean_text,
request.gemini_api_key, voice,
tts_audio_path, request.gemini_api_key,
tts_ass_path, tts_audio_path,
display_text=clean_text_for_display(request.text), tts_ass_path,
delay=2.0, display_text=clean_text_for_display(request.text),
) delay=2.0,
)
except Exception as gemini_err:
print(f"⚠️ Gemini TTS failed ({gemini_err}), falling back to Edge TTS")
await generate_tts_with_subs(
tts_clean_text,
voice,
tts_audio_path,
tts_ass_path,
display_text=clean_text_for_display(request.text),
delay=2.0,
)
else: else:
await generate_tts_with_subs( await generate_tts_with_subs(
tts_clean_text, tts_clean_text,
@@ -1323,17 +1335,28 @@ async def preview_tts(request: ReelRequest, x_api_key: str = Header(None)):
# Note: Gemini voices (fr-FR-Standard-A etc.) are valid Edge voices too, # Note: Gemini voices (fr-FR-Standard-A etc.) are valid Edge voices too,
# so we don't check for "Neural" - let edge_tts handle voice resolution # so we don't check for "Neural" - let edge_tts handle voice resolution
engine = request.tts_engine or "edge" engine = request.tts_engine or "gemini"
# Primary: Gemini TTS. Fallback: Edge TTS
if engine == "gemini" and request.gemini_api_key: if engine == "gemini" and request.gemini_api_key:
await generate_tts_gemini( try:
clean_text, await generate_tts_gemini(
voice, clean_text,
request.gemini_api_key, voice,
tts_audio_path, request.gemini_api_key,
tts_srt_path, tts_audio_path,
display_text=clean_text_for_display(request.text), tts_srt_path,
) display_text=clean_text_for_display(request.text),
)
except Exception as gemini_err:
print(f"⚠️ Gemini TTS failed ({gemini_err}), falling back to Edge TTS")
await generate_tts_with_subs(
clean_text,
voice,
tts_audio_path,
tts_srt_path,
display_text=clean_text_for_display(request.text),
)
else: else:
await generate_tts_with_subs( await generate_tts_with_subs(
clean_text, clean_text,