AwesomeTTS for Anki: Setup, Voices, and API Options (2026)
Updated August 2026. AwesomeTTS is still a useful Anki add-on if you want to generate audio from text, particularly when you want to bulk-create audio files, use a specific speech provider, or control exactly which field gets spoken.
The big change since I first wrote this guide is that text-to-speech has moved quickly. ElevenLabs is no longer a beta curiosity, cloud providers have changed their pricing and voice ranges, and Anki itself has more ways to play TTS without permanently generating an audio file. So I would no longer tell every learner to set up an Azure account on day one.
Contents
What is AwesomeTTS?
AwesomeTTS is an open-source Anki add-on that turns text in your cards into speech. You can use it to create audio files in bulk, generate audio while editing an individual card, or hear a word or sentence without recording yourself.
For language learning, the appeal is simple: I want to hear the target language while reviewing cards, but I do not want to record thousands of sentences manually.
If you’re new to Anki, start with our guide to using Anki for language learning. Get the card structure right first. Audio is an enhancement, not a substitute for useful cards.
Do you still need AwesomeTTS in 2026?
Sometimes. If all you want is for Anki to read a field aloud during review, a simpler built-in or template-based TTS setup may be enough. AwesomeTTS makes more sense when you want to save generated audio into the collection, process many notes at once, or choose among different cloud speech services.
I especially like saved audio when I want the same pronunciation every time and want the deck to work without making a new API request during every review.
The best TTS provider depends on the language
I used to recommend Microsoft Azure almost universally. Azure is still a strong option, but I would now test providers rather than assume there is one best engine for every language.
- Microsoft Azure Speech has a broad language and accent catalogue and works well for multilingual decks.
- Google Cloud Text-to-Speech has several voice families, with pricing and free allowances that vary by model.
- ElevenLabs now has a mature API and much broader multilingual support than it had when this article first mentioned it. Its pricing has also changed substantially.
- Language-specific providers can still sound better for a particular language. Do a listening test before generating thousands of files.
For dialect-heavy languages, synthetic speech still needs care. Arabic is the obvious example for me: a voice labelled “Arabic” may pronounce Modern Standard Arabic correctly while sounding wrong for the colloquial sentence you actually want to learn.
How I structure Anki cards with TTS
My basic language card still has three useful pieces of information:
- Native-language meaning
- Target-language sentence or phrase
- Target-language audio
I usually do not add native-language audio. I want my ears focused on the language I’m learning. I also avoid transliteration once I can read the native script, because I do not want to train myself to depend on Roman letters.
For example, the front might say “I really like eggplant,” and the back might contain Me encanta la berenjena plus the target-language audio. I also make the reverse card when recognition matters.
How to set up AwesomeTTS
Install AwesomeTTS from AnkiWeb, restart Anki if required, and first test it with a built-in or readily available service. Once you know the add-on is doing what you want, decide whether you need your own cloud API credentials.
- Choose the field you want spoken. Usually this is the target-language field.
- Test several voices. Listen for stress, vowel quality, numbers, names, and sentence rhythm rather than judging from a single word.
- Generate a small batch. Check filenames and playback before processing an entire deck.
- Only then connect your own API account if the included options do not give you the voice, quota, or control you need.
Using your own Azure or Google API key
The old version of this post described a specific Azure free allowance as though it were permanent. I would not build your workflow around a number copied from an old article. Cloud TTS pricing and free tiers change.
The general process is still the same: create an account with the speech provider, create a speech/TTS resource, generate credentials, note the required region or endpoint, and enter those details into AwesomeTTS. Set billing alerts if the provider can charge for usage, and never share your API key.
Google Cloud currently prices different TTS models differently and gives some models a monthly free allowance. ElevenLabs also changed API pricing in 2026. Check the provider’s own pricing page before generating a large deck rather than relying on the figures in a tutorial.
Where to get sentences
The TTS voice is only as useful as the sentence you feed it. I still prefer sentences from material I understand: textbooks, conversations, corrected writing, or examples I have checked with a speaker or teacher. AI tools can help create examples, but I verify anything that I plan to memorise.
My broader method is in the one-sentence-a-day language learning guide.
My recommendation
If you’re building a normal personal deck, start simple. Install AwesomeTTS, test the easiest available voice for your language, and only add an API account when you have a reason. If the pronunciation sounds wrong, change the provider rather than drilling bad audio because a particular service has the biggest feature list.
