Podcast Generation

1. Enter a topic or write your own script using Speaker 1:, Speaker 2:, etc.
2. Choose a model: Quickhorn (fast, free) or Silvermane Core / Studio / Studio Preview (Pro)
3. Select voices for each speaker and configure speakers (1-4)
4. Click Generate and wait for your podcast to be created
5. Singing tip: Add ♪ or (singing) only when you want a musical cue: "Speaker 1: ♪ Happy birthday! ♪"

Live Transcription

1. Click the microphone button or press and hold the space bar
2. Speak clearly into your microphone
3. Release to stop recording - transcription appears instantly
4. Copy or clear the transcript using the buttons below

Call & Meeting Recording

1. Upload an audio file (.wav, .mp3, .m4a, .flac, .ogg)
2. Click Transcribe to convert speech to text
3. Generate Summary to get key points, action items, and decisions
4. Download transcript (.txt) or summary (.md)

What's the difference between the TTS models?

Quickhorn (0.5B): Fast single-speaker voice model. Desktop runs it locally; web may use Unicorn's managed private inference backend. Best for quick drafts, TTS, and one-voice podcasts.
Silvermane Core (1.5B): Balanced quality and speed for regular production work. Requires Pro plan.
Silvermane Studio (Large): Best production-ready quality with the strongest overall voice output today. Uses more GPU (1.5x tokens). Requires Pro.
Silvermane Studio Preview (7B): Experimental next-gen voice preview. Promising, but not the primary production recommendation yet. Uses most GPU (2x tokens). Requires Pro.

How does pricing work?

Free ($0): 10 min podcast, 20 min transcription, 5 scripts/month
Starter ($14.99): 60 min podcast, 120 min transcription, 50 scripts/month
Pro ($29): 300 min podcast, 500 min transcription, 200 scripts/month + cloud TTS models

Note: Silvermane Studio uses 1.5x tokens, and Studio Preview uses 2x tokens. More speakers also increase usage.

Can I use my own API keys?

Yes! Go to Settings (click your profile) and add your own:
• Hugging Face Token - for cloud transcription fallback
• OpenAI API Key - optional for Whisper API transcription
• Ollama API Key - for cloud script generation fallback

You can still connect your own API keys in Settings if needed.

Why is my podcast generation slow?

Generation time depends on:
• Model choice: Silvermane Studio is the best production default; Studio Preview is experimental and may be slower
• Script length: Longer scripts take more time
• Number of speakers: More speakers = longer processing
• Cloud vs Local: Cloud models may sleep after inactivity (~30s wake time)

Tip: Use Quickhorn for the fastest turnaround. On desktop it runs locally; on web it may use Unicorn's managed private inference backend.

Where are my generated files stored?

All generated podcasts and TTS clips are automatically saved to your Library tab. You can:
• Play audio files directly in the browser
• Download files as WAV format
• Delete files you no longer need

Files are stored securely on the server and linked to your account.