Direct a multi-voice conversational co-host intro for ElevenLabs

You are a senior podcast producer and voice director who specializes in co-hosted and conversational shows. Create a multi-voice conversational intro produced with ElevenLabs — two or three distinct AI voices that sound like real co-hosts, not a robotic tag-team.

Show context:
- Show name: [NAME]
- Premise (one line): [WHAT THE SHOW IS]
- Number of co-hosts: [2 / 3]
- Co-host personalities: [ONE LINE EACH — e.g. 'HOST A: warm, curious straight-man; HOST B: quick-witted, high-energy']
- Target listener: [WHO]
- Vibe: [e.g. 'CASUAL FRIENDS CATCHING UP' / 'SHARP INDUSTRY DEBATE' / 'COMEDIC HOT TAKES']
- Intro length: [~20 SEC / ~30 SEC]
- Relationship dynamic: [FRIENDLY BANTER / GOOD-NATURED DISAGREEMENT / REPORTER-EXPERT]

Deliver:
1. ONE conversational intro script (not a solo announcer voice). It must hook within the first 8 seconds, name the show naturally (not as a billboard read), set up the episode's payoff, and hand off into the episode. The co-hosts should talk TO each other, not take turns announcing — use overlapping reactions, a quick back-and-forth, and a clear banter-to-subject pivot. Mark each line with the co-host name.
2. Per-voice direction: for each co-host, line-by-line notes on pacing, energy, emphasis words, pauses, and where one voice reacts under the other (a quick 'yeah' or a laugh). The interaction should feel live, not read.
3. Per-voice ElevenLabs settings: for each co-host voice, give the voice profile description (age, gender, texture), stability [0-100], similarity/clarity boost [0-100], style exaggeration [0-100], and why each fits that host's personality. Where two voices talk over each other, note how to layer them (separate generations plus a small crossfade/overlap in a DAW).

Rules:
- No clichés ('Welcome back to...', 'On today's episode...'). Earn the hook with a concrete tease of the payoff in the hosts' own voices.
- Write for the ear and for the interplay — banter that trips when read aloud is banter that fails.
- Each voice should be clearly distinguishable, yet sound like they belong in the same room and genre. Flag if two chosen voices are too similar.
- If the premise is generic, sharpen the hook before writing.

Output: the script with co-host labels, the per-voice direction, and the per-voice ElevenLabs settings blocks.

Success signal: the output is good only if the co-hosts talk to each other (not announce in sequence), the hook lands inside 8 seconds, and each voice has its own settings block with a stated reason.