Conversational AI Podcast Host Character
Budget: $30 – $250 AUD
I want to integrate a human-like, friendly, conversational AI co-host into my podcast so each episode feels like a natural dialogue rather than a rigid Q & A. The character should (1) open segments with pre-scripted questions I supply and (2) follow up in real time, adapting to guest responses with relevant probes or clarifications. A consistent vocal identity and light personality quirks—warm tone, occasional humor, no robotic clichés—are key to helping listeners connect.
You’re free to build with tools such as GPT-4, Rasa, or a custom Python stack, as long as the workflow lets me update question sets quickly and triggers the AI responses through my existing recording setup (Riverside.fm + Adobe Audition). Voice synthesis can rely on ElevenLabs, PlayHT, or a comparable engine; the final voice should be studio-quality and license-safe for commercial distribution.
Deliverables
• Character brief: background, speaking style, verbal tics, sample lines
• Voice model files (or service configuration) with clear usage instructions
• Script ingestion system that turns my question spreadsheets into prompts
• Real-time interaction logic with fallback phrases to handle unexpected answers
• Quick-start guide and live demo session to prove everything works in a podcast environment
Acceptance criteria
• AI reliably asks every pre-scripted question and produces follow-ups within 1.5 s latency
• Audio output matches loudness standards −16 LUFS stereo / −19 LUFS mono
• All code, prompts, and settings supplied so I can iterate without vendor lock-in
If the above is clear, let’s move straight to building and testing the first episode’s flow.
You’re free to build with tools such as GPT-4, Rasa, or a custom Python stack, as long as the workflow lets me update question sets quickly and triggers the AI responses through my existing recording setup (Riverside.fm + Adobe Audition). Voice synthesis can rely on ElevenLabs, PlayHT, or a comparable engine; the final voice should be studio-quality and license-safe for commercial distribution.
Deliverables
• Character brief: background, speaking style, verbal tics, sample lines
• Voice model files (or service configuration) with clear usage instructions
• Script ingestion system that turns my question spreadsheets into prompts
• Real-time interaction logic with fallback phrases to handle unexpected answers
• Quick-start guide and live demo session to prove everything works in a podcast environment
Acceptance criteria
• AI reliably asks every pre-scripted question and produces follow-ups within 1.5 s latency
• Audio output matches loudness standards −16 LUFS stereo / −19 LUFS mono
• All code, prompts, and settings supplied so I can iterate without vendor lock-in
If the above is clear, let’s move straight to building and testing the first episode’s flow.