Translation Engines
Every engine tried, top to bottom, until one succeeds. Drag to reorder — add a separator to stop the cascade at any point. Compare engines performance · Recommended cascade shape
Loading engine settings…
Recommended: Gemini. Live-benchmarked against every other engine here on the
same real content — Gemini (with a real-quota model) came out roughly
4–25× faster per subtitle cue than NVIDIA, Ollama, OpenRouter, or Groq.
The other engines are solid alternatives — useful as a fallback, or if you hit Gemini's daily cap.
Getting API keys
No engines configured yet — add one below.
⠿
— cascade stops here — anything below is never tried as a fallback —
⠿
{{ editBuffers[instance.id].name }}
{{ b }}
{{ editBuffers[instance.id].enabled ? 'enabled' : 'disabled' }}
rate-limited
auth failed
Shown in logs, History, and the Events tab — helpful if you add more than one instance of the same engine type.
Checking installed models…
Could not list installed models: {{ ollamaModelsError[instance.id] }}
Checking available models…
Could not reach /v1/models: {{ llamacppModelsError[instance.id] }}
● {{ testResults[instance.id].detail }}
● {{ pull.status }}{{ pull.total ? ' — ' + pull.pct + '%' : '' }}
● model ready
● {{ pull.error }}
● saved