How to use Azure Speech's free tier to read aloud in Lullula
Microsoft's Azure Speech has a Free (F0) tier that, as of October 2026, includes 500,000 characters of neural text-to-speech every month. Connect it in Lullula with your own key and Azure's voices read web pages in the same player as ours: reading starts at the article's first sentence, and each sentence lights up as it's spoken. That's about 80,000 English words, or roughly 9 hours of listening, a month, and a Free-tier resource never runs up a bill.
Setup takes about 15 minutes: create an Azure account, create a free Speech resource, copy its key and region, and connect it in Lullula. This guide walks through each step, what counts toward the allowance, and what each error message means.
Availability: Voice providers are in Lullula 1.8 and later, on desktop Chrome, Edge and Firefox.
Is this for you?
It is if you listen a lot and don't mind a one-time setup in Microsoft's cloud console, which asks for a card to confirm who you are. It isn't if you just want natural voices with no setup. Lullula's own voices need none of this: a free Lullula account adds about 2,000 premium words each month, shared with AI summaries, and Pro covers 200,000 words a month.
What the free tier gives you
Azure lists 0.5 million neural characters a month among its always-free services, so it isn't a trial: it renews every month for as long as your account is active. Microsoft can change it, and its Azure Speech pricing page is the authority.
| Free (F0) tier | |
|---|---|
| Neural voice allowance | 500,000 characters a month |
| In English | About 80,000 words, roughly 9 hours of listening (at six characters a word, spaces and punctuation included) |
| In Chinese or Japanese | Each Chinese character, kanji included, counts as two, so about 250,000 characters |
| What counts | Letters, numbers, spaces and punctuation |
| Request rate | 20 requests per 60 seconds, not adjustable |
| When it runs out | Azure refuses requests until the allowance resets; nothing is charged |
| How many | One free Speech resource per subscription |
The allowance is listed for Azure's standard neural voices. HD voices are priced separately, so if one shows up in your list, don't assume it's covered. Microsoft explains what it counts in billable characters and sets the request rate in Azure Speech quotas and limits.
Before you start
You need four things, and about 15 minutes:
- A Microsoft account or a GitHub account to sign in to Azure.
- A phone number for a verification code.
- A credit or debit card (not prepaid). Azure uses it to check you're a real person and may place a temporary $1 hold that it reverses.
- Lullula 1.8 or later in desktop Chrome, Edge or Firefox.
You don't need a Lullula account for Azure's voices. Azure's menus change often; the paths below were right when we wrote this, and Microsoft's own pages, linked in each step, are the authority if a button has moved.
Step 1: Create an Azure account
Open Azure's account page. It offers two kinds of account, and both include the free Speech allowance:
| Try Azure for free | Pay as you go | |
|---|---|---|
| Free monthly Speech allowance | Yes | Yes |
| Extra | $200 credit for 30 days | None |
| After 30 days | You must move to pay as you go, or the account and its services are disabled | Nothing to do |
| Card charged | Never, unless you move to pay as you go | Only for usage beyond free amounts |
The catch with the free account: on day 30, your Azure key stops working in Lullula unless you move to pay as you go. Moving doesn't make your free Speech resource cost anything. A Free (F0) resource has no paid overage; you're billed only if you create or switch to a paid Standard (S0) resource.
To sign up:
- Choose Try Azure for free or Pay as you go.
- Sign in with your Microsoft or GitHub account, or create one.
- Fill in your profile. For Country/Region, pick the one that matches your card's billing address.
- Verify your phone number with the code Azure sends.
- Add your card and accept the agreement.
When it finishes, you land in the Azure portal.
Step 2: Create a Speech resource on the Free tier
A resource is Azure's word for one copy of a service with its own keys. You want a standalone Speech resource, because that's the kind with a Free (F0) tier.
- In the portal, choose Create a resource.
- Search for Speech, select Microsoft's Speech service (Microsoft also calls it Azure Speech in Foundry Tools), and choose Create. Don't pick a Microsoft Foundry resource: it's billed at Standard rates and doesn't include the free tier.
-
On the Basics tab, fill in:
- Subscription: the one you just created.
- Resource group: choose Create new and name it, for example lullula-voices. It's just a folder.
- Region: one close to you, such as East US or West Europe. Note it; Lullula needs it later.
- Name: anything unique, for example lullula-yourname.
- Pricing tier: Free F0.
- Leave the other tabs as they are. If you open Network, keep access open to all networks, because Lullula calls Azure from your browser.
- Choose Review + create, then Create. It takes under a minute.
- Choose Go to resource.
If Free F0 isn't offered, this subscription already has a free Speech resource. Azure allows one, so use that one instead.
Step 3: Copy your key and region
In your new resource, open Keys and Endpoint under Resource Management in the left menu. It shows KEY 1, KEY 2 and Location/Region.
- Copy KEY 1. KEY 2 works just as well; there are two so you can replace one without downtime.
- Note the Location/Region, for example eastus.
Treat the key like a password: anyone who has it can use your allowance. If it ever leaks, regenerate it on the same page and the old key stops working.
Step 4: Connect Azure in Lullula
- Click the Lullula icon in your browser's toolbar and choose Settings.
- Open Voice providers, the section right after Voices. If you don't see it, update the extension.
- On the Azure Speech card, paste your key into Subscription key.
- Type your Region. Lullula accepts it the way the portal shows it (East US) or as its id (eastus).
- Press Connect.
Lullula checks the key by asking Azure for its voice list. When that works, the card says Connected · N voices · refreshed just now. When it doesn't, the error appears in red under the fields and nothing is saved.
Step 5: Choose an Azure voice and press play
Connecting doesn't change the voice you hear; you pick one. There are two ways:
- In Settings: go to Voices and set the Plan filter to Own key. Every voice Azure offers in your region appears, grouped by language, with the provider's name on each card.
- In the player: open the voice picker on any page. A Mine tab appears next to Premium and Free when Azure has a voice for that page's language.
Each Azure voice belongs to one language, so choosing one changes the voice for that language only. Set one for each language you listen in, then open an article and press play. Previewing a voice in Settings is a real request to Azure, so it uses a few characters of your allowance.
Keep an eye on your allowance
Azure shows exactly how much you've used. In your Speech resource, open Metrics under Monitoring, choose the Synthesized Characters metric with Sum, and set the time range to this month. Under Alerts you can also ask for an email when usage passes, say, 400,000 characters (Microsoft's metrics guide).
What counts toward the 500,000:
- Every sentence you play. A 2,000-word article is 10,000 to 12,000 characters.
- A few sentences ahead. To avoid gaps, Lullula prepares up to four sentences ahead of the one you're hearing (at most two requests at a time). Stop halfway and those few still count.
- Replays. Audio isn't kept after you close the page, so listening again uses characters again.
- Previews of voices in Settings.
When the month's allowance is gone, Azure refuses further requests and the player says Azure refused the request: quota or rate limit. Nothing is charged. Switch to Lullula's voices or your browser's built-in voices until the allowance resets.
Where your key and text go
Your key stays in the browser profile where you connected it. It isn't synced to other devices and is never sent to Lullula. When you play a page with an Azure voice, the requests that make the audio go straight from your browser to Azure, and Microsoft's own terms apply to that text.
Connecting Azure changes only who makes the audio. AI summaries send the page's text to our servers and an AI provider, and language detection, the page outline and, if you're signed in, your reading history still reach Lullula, whichever voice reads. Our privacy policy has the details. To disconnect, press Remove on the Azure Speech card; it deletes the key from this browser straight away.
Troubleshooting
| What you see | What it means and what to do |
|---|---|
| Azure rejected the subscription key for region …; check the key and that the region matches the resource (on Connect) | The key and region come from different resources, or the key was copied with a space or cut short. Copy both again from the same Keys and Endpoint page. |
| a region such as eastus is required | The Region field is empty or not a region. Type it as the portal shows it. |
| Your Azure key was rejected. Check it in Lullula settings. (while reading) | The key was regenerated or the resource deleted. If it starts about 30 days after you signed up for a free account, the account has likely been disabled: move it to pay as you go in the portal. |
| Azure refused the request: quota or rate limit. | Either the month's 500,000 characters are used up, or you passed 20 requests a minute. Each sentence is a request, so the rate limit is likeliest at high speed on pages with many short sentences. Lullula retries once; if it keeps happening, slow down a step or wait a minute. |
| No Mine tab in the player | Azure has no voice for this page's language. An Azure voice for Spanish won't show on an English page. |
| A different voice on another computer | The key isn't synced. Connect Azure on that computer too, with the same key. |
For other messages, see the error list in our voice provider guide, or email support@lullula.io.
Frequently asked questions
Will Azure ever charge me for this?
Not for a Free (F0) Speech resource. With a free account, Azure says it never charges your card unless you move to pay as you go. On pay as you go, you're billed only for paid resources or usage beyond free amounts, and a Free-tier resource simply stops at its limit.
Can I get more than 500,000 characters a month?
Yes, by switching the resource to the Standard (S0) tier on its Pricing tier page. From then on Azure bills you at its neural voice rate, listed on its pricing page. Lullula Pro is the flat-price alternative, at 200,000 words a month.
Which Azure region should I choose?
One close to you, so audio starts sooner. Lullula lists every voice Azure offers in that region.
Do Azure voices work in the Lullula web app or in Listen later?
No. They play only in the browser extension where you connected Azure. The web app and the audio Listen later prepares in advance use Lullula's own voices.
Do Azure voices work on PDFs, EPUBs and Google Docs?
They go through the same player that reads PDFs, EPUBs and Google Docs, so they should. We haven't yet tested provider voices on those formats. If something doesn't work, email support@lullula.io.
Azure's voices or ours, one player
Connecting Azure doesn't replace anything. Lullula's own voices stay in the list, and you can switch between them and Azure's from the player at any time, which helps when the month's allowance runs low. Using another provider? Our voice provider guide covers ElevenLabs, OpenAI (or an OpenAI-compatible service), Azure Speech, Google Cloud Text-to-Speech, Amazon Polly and Alibaba Cloud Qwen-TTS.