Russian dialogue audio for Anki cards
Click a button on a Russian flashcard and this service generates a short
pedagogical dialogue that uses the word, voices it with two Russian speakers,
and attaches the MP3 directly to the card. The point is to hear the word
in living context — not as a recited isolated lemma.
How it works
1 · Anki
Click "New Audio"
on a card in review or in Browse
→
2 · This server
/generate-card
FastAPI on DigitalOcean VPS
→
3 · Claude
Writes the scene
Sonnet 4.6, ~3 short dialogues
→
4 · Gemini TTS
Voices the scene
multi-speaker Russian
→
5 · Merge
Stitch + piano gap
WAV → MP3
→
6 · Anki
Saved to card
embedded as [sound:…]
The stack, in plain English
- Anki add-on — a Python plugin that runs inside Anki on the Mac.
Adds two buttons ("New Audio", "Custom Audio") plus a Browse menu for
batch jobs.
- This server — Python
FastAPI app on a DigitalOcean
VPS. It talks to Claude and Gemini, merges the audio, sends the MP3 back.
- Cloudflare — fronts the server with HTTPS at
ankiaudio.fabianhagen.com.
- Claude (Anthropic) — writes a short multi-scene dialogue around
the target word, with explicit voice/stage directions.
- Gemini TTS — multi-speaker Russian voices (24 kHz WAV).
- SQLite — logs every card. Audio files live on disk for 45 days
so you can replay them on the logs page.
View past generations and add manual practice clips on the
logs page.