Playgrounds

The Playgrounds are isolated test surfaces — pick a service, send it some input, see the output. No character, no scenario, no chat history to muddy the result. Use them when you want to evaluate a single service in isolation.

Transcription Playground

Test your Speech-to-Text service. Speak into your mic and see the transcription appear in real time. Useful for:

• Checking that your STT service is wired up correctly.
• A/B comparing services (set one as active, transcribe a phrase, swap, transcribe again).
• Tuning microphone gain and noise.

Voice Playground

Test Text-to-Speech. Type text, pick a voice, hear it back. Useful for:

• Auditioning voices for a character.
• Testing emotion / multilingual settings (ElevenLabs especially).
• Comparing TTS services on the same phrase.

Text Generation Playground

Test your LLM directly with system/user/assistant message construction. Useful for:

• Prompt engineering — iterate on a system prompt without a character involved.
• Measuring response latency and token throughput.
• Quick "is this model working at all" sanity check after configuring it.

The playground reports speed metrics and token counts for each generation so you can compare models on the same prompt.

What's next