docs(vapi-voices): list every active voice, not just the documented subset

[Mehayla]

Complete list of active voices

Adds a consolidated reference table to the Vapi Voices page listing every currently active voiceId (all voices that are not part of the March 1, 2026 legacy retirement), including the eight continuing voices (Leo, Zoe, Mia, Jess, Zac, Dan, Leah, Tara) that previously had no entry on this page at all.

• New "Complete list of active voices" table under Active Voices, covering all 21 current voices with their category (Continuing / New, V2 or not) and whether a full Gender/Accent/Age/Characteristics writeup exists below.
• For the 8 continuing voices without a detailed writeup yet, the table links to the Voice Library to preview them, and a <Note> cross-links the legacy voice migration guide for the full retirement/replacement mapping, instead of leaving them undocumented.
• No other sections changed; all existing per-voice detail blocks, language/multilingual support content, and setup instructions are unchanged.

[Mehayla]

Why this gap, and why now

Source population: Plain Support Tickets, Fern Ask AI Conversations, Slack, Gong, Discord, and Posthog NPS (all customer-facing feedback sources in the graph).

Volume & trend: Over a 4-week window (Aug 15 - Sep 12, 2026, anchored to the latest complete Ask AI Conversations coverage rather than wall-clock "today" — see coverage caveat below), the Voice Providers topic ran ~130 conversations/week (current week) vs. ~139 (prior week) vs. an ~111/week trailing 4-week average — a sustained, above-baseline topic rather than a one-off spike. Within it, Vapi Built-in Voices was the single largest sub-topic (90 of the ~316 four-week Voice Providers conversations).

Theme breakdown inside "Vapi Built-in Voices" (4 weeks):

• Finding The Current Built-in Voice Lineup — 22
• Help With Voice Capability Documentation Gaps — 19
• Need Clearer Information On Voice Availability — 11
• Integrating Vapi Voices In Apps (api/dashboard) — 10
• Built-in Voice Catalog Changes — 8

These five themes (~70 conversations) all cluster around one question: "what Vapi voices exist right now, and are they still active?" Representative Ask AI transcripts show the bot repeatedly reconstructing a voice list from scratch, and one Gong call recording a customer's concern that "vosotros solo tenéis voces americanas" ("you only have American voices") — i.e. uncertainty about what's actually in the catalog.

Ask AI coverage signal: The outcome field is populated on 100% of Ask AI Conversation records in the org (24,039/24,039), with source coverage running 2026-04-01 through 2026-09-11. Within Vapi Built-in Voices, Ask AI outcomes over the 4-week window were 296 answered, 17 refused, 3 silent — mostly answerable from existing docs, but recurring at this volume regardless. Caveat: Ask AI Conversations data lags ~10 days behind wall-clock ("today" is 2026-09-21; latest Ask AI record is 2026-09-11), so the "current week" window in this analysis is anchored to 2026-09-05–09-12, the latest fully-covered week, not the trailing 7 days from today.

Live-doc audit finding: fern/providers/voice/vapi-voices.mdx already documents Vapi Voices V2, per-voice language/multilingual behavior, and full Gender/Accent/Age/Characteristics detail for 13 voices (Elliot, Savannah, Rohan, Emma, Clara, Nico, Kai, Sagar, Godfrey, Neil, Layla, Sid, Naina). Separately, fern/providers/voice/vapi-voices/legacy-migration.mdx states that Leo, Zoe, Mia, Jess, Zac, Dan, Leah, and Tara are also continuing/active voices — but those 8 voices have zero entry on the canonical Vapi Voices reference page. That's the concrete gap: there is no single place documenting the complete current voice catalog, which matches exactly what customers and Ask AI keep asking for.

What the PR does

Adds a "Complete list of active voices" table to vapi-voices.mdx enumerating all 21 currently active voices (11 continuing + 10 new V2 clones) with their category and whether a full writeup exists below. For the 8 continuing voices that still lack per-voice Gender/Accent/Age/Characteristics detail, the table links to the Voice Library for a live preview rather than fabricating attributes we don't have sourced data for, and cross-links the legacy migration guide for the full retirement/replacement mapping.

Deliberately out of scope / follow-up

• Not done: Adding full Gender/Accent/Age/Characteristics/sample-audio entries for Leo, Zoe, Mia, Jess, Zac, Dan, Leah, Tara. We don't have a verified source for those attributes, so we linked to the Voice Library instead of guessing. A follow-up PR could add these once source data is available.
• Not done: Documenting a Voice Library API for filtering by language/accent/gender (theme "Integrating Vapi Voices In Apps", 10 conversations). The live Voice Library page confirms filtering today is dashboard-only (Provider + name search) — this looks like a product capability gap, not a documentation gap, so it's flagged here rather than addressed in this PR.
• Not done: Third-party provider language/accent coverage questions (Russian on ElevenLabs/Cartesia, Telugu, Estonian) surfaced in the same Voice Providers topic are about non-Vapi TTS providers and are out of scope for the Vapi Voices page.
• No navigation change: fern/docs.yml navigation entry for this page is unchanged; this PR only edits page content.

Evidence coverage caveat

This analysis covers customer-facing feedback sources only (Plain Support Tickets, Ask AI Conversations, Slack, Gong, Discord, Posthog NPS) and is anchored to the latest fully-covered week (2026-09-05 to 2026-09-12) because Ask AI Conversations data lags ~10 days behind wall-clock time.