Gradium

@Gradium list all available voices in English and Spanish, then show me which ones support emotional tone control.

Found 47 English voices and 23 Spanish voices in the Gradium catalog. Of these, 18 English voices (including 'en-US-neural-aria', 'en-GB-neural-sonia') and 9 Spanish voices ('es-ES-neural-elvira', 'es-MX-neural-dalia') support emotional tone parameters like 'cheerful', 'empathetic', and 'calm'. Each voice entry includes sample audio URLs, language codes, and supported style tags. Custom voices for your organization are also listed separately if configured.

This example demonstrates catalog discovery — useful when selecting voices for text-to-speech workflows. The MCP returns metadata only; you'll need separate API calls to generate actual audio. Voice availability depends on your Gradium subscription tier.

@Gradium create a pronunciation dictionary for en-US that rewrites 'SQL' as 'sequel' and 'API' as 'A P I' spelled out.

Created pronunciation dictionary 'en-US-tech-terms' (UID: dict_8x3kp2m) with 2 rewrite rules. The dictionary is now active for all en-US voice synthesis requests tied to your API key. When generating audio, 'SQL' will be pronounced as 'sequel' and 'API' will be spelled letter-by-letter. You can attach this dictionary to synthesis calls by referencing its UID or update rules anytime via the MCP.

This write operation modifies your Gradium account state immediately. Pronunciation dictionaries persist across sessions and affect all future synthesis requests unless explicitly detached. Useful for brand names, acronyms, or domain-specific jargon that default TTS engines mispronounce.

@Gradium check my current credit balance and summarize this month's usage — how many audio minutes have I generated and what's my burn rate?

Your Gradium account has 12,450 credits remaining (out of 25,000 monthly). This billing period (Jan 1–31), you've consumed 12,550 credits across 487 synthesis sessions, generating 3.2 hours of audio. Average burn rate: ~450 credits/day. At this pace, you'll exhaust your balance in ~28 days. Top voice used: 'en-US-neural-aria' (38% of sessions). No overages yet, but you're trending toward your cap before month-end.

This synthesis example pairs Gradium's usage and credit APIs with the AI's math reasoning to forecast spend. Requires API_KEY auth with billing-read scope. Credit costs vary by voice type (neural voices cost more than standard), so burn-rate projections are estimates based on historical mix.

When Gradium fits voice-heavy support teams under 10 seats

A 6-person support team records customer calls and needs accurate transcripts with custom pronunciation for product names and industry jargon. Gradium's pronunciation dictionaries let you define how "Gradium" or "OAuth" should render in text, then apply those rules across every transcript. The Get Usage Summary tool tracks credit burn per billing period, so you know when you're approaching your plan cap. This works cleanly if your call volume stays under 200 hours per month and you're willing to manage dictionaries manually. Above that threshold, the 9-tool surface area feels thin—you'll want batch-upload APIs and webhook triggers that aren't exposed here. If your team is already in Gradium's ecosystem and needs light customization, the MCP saves you from building your own pronunciation layer.

Gradium MCP handles voice catalog browsing, not editing

A 3-person podcast team uses Gradium to generate intro voiceovers and ad reads in multiple languages. The List Voices tool surfaces the full catalog with metadata filters, so your AI agent can pick the right voice for a Spanish ad spot without opening the web UI. Get Voice pulls detailed metadata for final selection. The limitation: this MCP doesn't expose audio synthesis or editing endpoints—you're browsing and configuring, not rendering. If your workflow is "pick a voice, then hand off to Gradium's web app or a separate API client," the MCP fits. If you want the AI to generate and iterate on audio files inside the same session, you'll hit the boundary fast. Best for teams that treat voice selection as a research step, not a production step.

When pronunciation dictionaries justify the API-key setup cost

A 12-person product team ships onboarding videos in 8 languages and needs consistent pronunciation of the product name, feature labels, and competitor terms across every voiceover. Gradium's Create Pronunciation Dictionary tool lets you define rewrite rules per language, then retrieve and update them programmatically. The Get Credits tool prevents mid-sprint surprises when your video editor burns through the monthly allowance. This setup makes sense if you're producing 20+ videos per quarter and pronunciation errors have caused customer confusion in the past. If you're only localizing 3 videos a year, the overhead of managing API keys and dictionaries outweighs the benefit—just use the web UI. The MCP pays off when pronunciation consistency is a repeatable, high-frequency problem.