Setup · every provider

Get your API keys

You don’t need any key to start: Reverbot’s own AI answers your included questions. A key of your own lets a feature use your account at that company, which bills you directly — and Gemini’s is free within Google’s limits. Every key field in the app has How to get a key with the same steps as this page, and a link straight to where the key is made. Keys stay in your iPhone’s Keychain and go only to their own company.

Get a Gemini key (free tier) ↗

Jump to: Gemini · ChatGPT · Claude · Grok · Google Cloud voices · ElevenLabs · Anam & Simli · Respeecher & Resemble · BytePlus · Twelve Labs · Cloudflare

AI answers

One key answers your questions. Gemini’s is free within Google’s limits; the others bill you directly.

Gemini (Google)

Free tier

The quickest way to keep Reverbot answering at no cost: free within the limits Google sets for your project, which AI Studio shows. A Google account is all you need — no billing to start. The same key also works for the Gemini voices, the live rehearsal partner on Gemini, and Google’s pictures and videos in Image & Video AI.

Get your key ready during the free trial. Tap Get your key — required later beside the question meter → Set up a free Gemini key. At zero free questions the link says Get your key — required now: your own AI key is needed to keep answering, even if the trial is still active. Settings → AI & API Keys accepts a key at any time.

  1. Tap Get my Gemini key in Reverbot, or go to aistudio.google.com/apikey. Sign in with any Google account, and accept Google’s terms if it asks — an app can’t accept them for you.
  2. New to AI Studio? Google has already made a project and a key for you: they’re on the API Keys page. Otherwise tap Create API key and choose a project, or Create project. Already use Google Cloud? Import projects (Dashboard → Projects) brings yours in.
  3. Copy the key with the copy icon (two overlapping squares) on its row. Keys are shown shortened, like …QWaA.
  4. Back in Reverbot, tap Paste. That saves it and switches answers to Gemini.
An older key stopped working? Google now refuses keys that aren’t restricted to the Gemini API. Every new key from AI Studio already is, so the simplest fix is a new key.
Free, with limits. Google caps requests per minute and per day on the free tier, renewing daily; it no longer publishes the numbers and shows your key’s own in AI Studio. If you hit a limit, Reverbot tells you and offers to switch. Free means Google may look — unless you are in the European Economic Area, Switzerland or the UK: under Google’s unpaid terms, what you send on a free key can be read by human reviewers and used to improve its products (Google’s terms). Turning billing on in AI Studio moves the key to the paid terms everywhere.
Open Google AI Studio ↗

ChatGPT (OpenAI)

No free API tier, but short spoken answers cost little. The same key powers the OpenAI voices, the live rehearsal partner on OpenAI, and OpenAI pictures in Image & Video AI. A ChatGPT subscription doesn’t pay for API use.

  1. Go to platform.openai.com/api-keys and sign in (a ChatGPT login works there).
  2. Add a payment method or prepaid credit under Settings → Billing. You can set a monthly limit under Limits.
  3. On API keys, tap Create new secret key, name it (Reverbot, say) and copy it. It starts with sk-, and OpenAI shows it only once.
  4. In Reverbot: Settings → AI & API Keys → Advanced, tap the sliders beside ChatGPT, paste the key, then go back and tap ChatGPT’s row to use it.
Open OpenAI API keys ↗

Claude (Anthropic)

No free API tier — it runs on prepaid credits. A Claude Pro or Max subscription doesn’t include API use.

  1. Go to the Claude Console: platform.claude.com/settings/keys, and sign in.
  2. Buy credits under Billing. You can set a spend limit under Limits.
  3. Go to API keys → Create Key, name it and copy it. It starts with sk-ant-, and it’s shown only once.
  4. In Reverbot: Settings → AI & API Keys → Advanced, tap the sliders beside Claude, paste the key, then go back and tap Claude’s row to use it.
Open the Claude Console ↗

Grok (xAI)

No free API tier — it runs on prepaid xAI credits. The same key powers the Grok voices and Grok Imagine pictures and videos in Image & Video AI.

  1. Go to the xAI Console: console.x.ai, and sign in.
  2. Add credits under Billing.
  3. Go to API Keys → Create API Key, name it and copy it. It starts with xai-.
  4. In Reverbot: Settings → AI & API Keys → Advanced, tap the sliders beside Grok, paste the key, then go back and tap Grok’s row to use it.
Open the xAI Console ↗

Voices

Reverbot’s own voice needs no key. These give you more voices on your own account.

Google Cloud voices

Google’s voices at any quality tier, on your own Google Cloud project. This key is only for voices; AI answers use a Gemini key.

  1. Open Google Cloud’s Text-to-Speech page, sign in with any Google account and create a project (any name).
  2. Turn on billing for the project: Billing → Link a billing account. Google needs a payment method even for the free characters, and charges only past them.
  3. Back on the Text-to-Speech page, tap Enable.
  4. Make a key: APIs & Services → Credentials → Create credentials → API key.
  5. Lock it down: open the key, choose API restrictions → Restrict key → Cloud Text-to-Speech API (and, if you like, Application restrictions → iOS apps → ai.reverbot.app).
  6. In Reverbot: Settings → Voice & Speed, choose Google Cloud, paste the key, pick a tier and a voice, and tap Test voice. Optional: Billing → Budgets & alerts in Google Cloud emails you as you near a monthly budget.
Open Google Cloud ↗

ElevenLabs

Free plan, then paid

The most human-sounding voices, a Soul’s own voice, and Voice Changer conversion — one key for all three. The free plan has a monthly allowance; making a voice of your own needs a paid plan.

  1. Go to elevenlabs.io/app/developers/api-keys and sign in. A free account works.
  2. Tap Create API key and name it. If you restrict what it may do, allow Text to Speech, Speech to Speech and Voices (read and write): Reverbot uses those.
  3. Copy the key — ElevenLabs shows it only once.
  4. In Reverbot, paste it in Settings → Voice & Speed (ElevenLabs), on a Soul’s voice page, or in Voice Changer: it’s the same key everywhere.
Open ElevenLabs API keys ↗

Faces, voices you convert, pictures and videos

All optional, each on your own account, and each says what it costs before anything is sent.

Anam and Simli: a realistic talking face

Your recorded face talks for free on the phone. For a realistic face, choose a paid animation on your own account under Settings → Center button appearance → Personal face → How the face talks, and paste its key there. Each one shows its cost and terms before anything is uploaded.

  1. Anam (a live face, billed by the second while it’s connected): go to lab.anam.ai/api-keys, sign in, tap Create API key, and copy it.
  2. Simli (a clip of each sentence, using your Simli credits): go to app.simli.com/apikey, sign in, and copy the key shown.
  3. Paste it in Reverbot and tap Save Anam key or Save Simli key.

Respeecher and Resemble: Voice Changer conversion

Voice Changer’s effects are free and run on the phone. To convert a recording into a natural AI voice, choose a company under Settings → Voice Changer → Natural AI voices and paste its key there (ElevenLabs works too — see above). Each conversion is charged to your own account, after a confirmation that states the price.

  1. Respeecher: go to marketplace.respeecher.com/account, sign in to Respeecher Marketplace, and generate an API key with an expiry date. A key from Respeecher’s text-to-speech playground (Space) won’t work. When it expires, make a new one.
  2. Resemble: go to app.resemble.ai, sign in, open Account → API, and copy your API token. Resemble also needs your own Cloudflare storage (below).
  3. Paste the key in Reverbot and tap Save key.

BytePlus: Seedream pictures and Seedance videos

BytePlus (ByteDance’s cloud) makes Seedream pictures and Seedance videos, including a Soul’s look and its world. The Gemini, xAI and OpenAI keys above work for pictures and videos too, in Settings → Image & Video AI. Every request says what it costs first.

  1. Go to the BytePlus console and sign up; BytePlus asks you to verify your identity.
  2. In ModelArk → Model activation, turn on Seedream for pictures and Seedance for videos.
  3. Open API Keys → Create API Key and copy it.
  4. In Reverbot, paste it in Settings → Image & Video AI (or on a Soul’s look or world page — it’s the same key) and choose the region it was made in. Seedance needs an Asia Pacific key, and BytePlus opens Seedance 2 once your account holds USD 30 of credit.

Twelve Labs: Souls watching whole videos

Free plan, then paid

Twelve Labs lets a Soul watch a whole video, picture and sound. It bills your own account by the minute of video, and Reverbot shows the cost before each one. Without it, Gemini on your key watches shorter videos whole.

  1. Go to playground.twelvelabs.io and sign up — it’s free.
  2. On API Keys, tap Create API Key and copy it.
  3. In Reverbot, paste it in Settings → AI & API Keys, under Video AI — Twelve Labs, or on a Soul’s photos page.

Cloudflare: video calls, recordings and Soul media

Free allowance, then paid

Video calls and meetings, recordings and a Soul’s pictures and videos run on your own Cloudflare account. Open Chat settings → Video calls & recordings and follow its guided setup: it opens Cloudflare with the right token settings already filled in, and you paste the token it makes. Cloudflare bills you directly past its free allowance.

Which model should you pick?

Every provider offers a few. Reverbot rates each one for speed, and recommends the fastest — here is why that is the right trade for a spoken assistant.

  1. Fastest ●●● — answers begin almost immediately. This is the one marked Recommended, and the one you get unless you change it.
  2. Fast ●●○ — a short pause before it starts talking.
  3. Slower ●○○ — a noticeable pause before it starts talking. Usually the provider's most capable model.

You'll find the rating beside every model in Settings → AI & API Keys — tap a provider, then Model — so you can see what a change will cost you before you make it.

Why the fastest, and not the cleverest? Reverbot answers out loud, in your ear, usually in a sentence or two. The larger models are built to think before they answer — which is worth waiting for in an essay, and worth very little in one spoken sentence. What you would actually notice is the silence: you stop talking, and nothing happens for a while. Reverbot starts speaking as soon as the first sentence is ready rather than waiting for the whole answer, so most of the difference is hidden — but the wait before that first sentence is not, and that is what the speed rating measures.

Answer speed

Most AI models now think before they answer — and several do it unless you tell them not to. Reverbot asks them to think as little as they allow.

  1. Fastest — the default. The least thinking each AI allows — none at all where it can be switched off — so the answer starts as soon as possible. Right for answers you listen to.
  2. Thoughtful — lets the model think first. Slower to start, and it can be better on a genuinely hard question.

The switch is in Settings → AI & API Keys, under Answer speed, and it applies to whichever provider you are using.

Why this is on by default. A thinking step is worth waiting for in a written answer you can skim. Reverbot's answers are a sentence or two, spoken into your ear, so the thinking mostly buys you silence — and on some providers you are billed for it as well. Keeping it to the least each AI allows is the single biggest difference you can make to how quickly Reverbot starts talking.
Paid fast modes. Two providers sell a faster lane, and Reverbot offers each one only while you are actually using that provider. Both are off unless you switch them on, because both bill at roughly double.

OpenAI fast mode — priority processing, up to 2.5× faster. Nothing to sign up for, and if it isn't available for a request OpenAI serves that one at normal speed rather than failing it.

Claude fast mode — appears on Claude Opus, and needs fast-mode access on your Anthropic account. It streams up to 2.5× more words per second, which speeds up the rest of a long answer rather than the wait before the first word. Without access, Reverbot falls back to normal speed on its own.

Hit Gemini’s free-tier limit?

Reverbot watches for it and lets you know the moment it happens — you’re never left wondering why an answer didn’t come. Here is what to do about it.

  1. Wait it out — per-minute limits reset in seconds; the daily allowance comes back in full the next day.
  2. Upgrade Gemini — enable billing in Google AI Studio for much higher limits (you still only pay for what you use).
  3. Switch providers — add an OpenAI, Claude or Grok key (Settings → AI & API Keys → Advanced, the sliders beside it), then tap its row to use it for answers.
Your keys stay yours. Keys live in your iPhone’s Keychain and are only ever sent to the provider you choose. Reverbot never sees or stores them on a server.
How to use Reverbot →