FAQ
Not here? Ask on GitHub.
Is SubCast free?
Yes: free, open source, MIT license. AI chat, guides and translation use your own model key, billed by the provider you pick. Without a key you can still subscribe, play and read existing subtitles and deep reads.
What does the AI cost?
Translation uses free Google Translate by default. For chat and guides, DeepSeek is the recommended model and costs roughly one jiao (about a US cent) per long episode. You can also point SubCast at local models in Ollama or LM Studio and pay nothing at all.
Is my data uploaded?
API keys live in the system keychain. With on-device transcription, audio never leaves the machine. Cloud transcription, AI chat and translation send the audio or text they need to the provider you choose. Subtitles you transcribe are shared anonymously with a public library by default, text only, and you can turn that off (see the next question). Play history and downloads stay on your machine.
Does it work in mainland China?
Podcasts, Xiaoyuzhou and the Qiaomu deep reads work directly. YouTube needs a proxy, or any network that can reach YouTube. When Google Translate is unreachable, the app falls back to a domestic relay automatically.
Windows says the installer is risky. What now?
The Windows installer has no paid code signature, so SmartScreen shows a warning. Choose "More info", then "Run anyway". The macOS build is signed and notarized by Apple and shows no such warning.
How do updates work?
They are automatic. The app checks about 10 seconds after launch, then every 6 hours. The update manifest is signed and verified before installation, and the app asks you to restart when it is ready.
Which systems are supported?
Installers exist for macOS (Apple silicon and Intel) and 64-bit Windows 10/11. Linux has AppImage and deb builds with less testing. The interface comes in Simplified Chinese, Traditional Chinese, English and Japanese.
What can I subscribe to?
Public podcasts from around the world, found by name through the Apple Podcasts directory. Also YouTube channels, playlists and single videos, Xiaoyuzhou, any RSS feed, and the Qiaomu deep-read channels. The Discover page has about 60 curated shows to start with.
What about videos without captions?
SubCast transcribes on your device: it downloads the Qwen3-ASR model (about 1GB) on first use, and audio never leaves the machine. Cloud transcription services are supported too, billed by the provider.
Which AI models are supported?
DeepSeek, SiliconFlow, Kimi, Claude, OpenAI, OpenRouter and Gemini, plus local models in Ollama and LM Studio, and any OpenAI-compatible endpoint. Translation defaults to free Google Translate and can be switched to your own model.
How do I report a problem?
Open an issue on GitHub.