Full guide
What a Google AI Studio audiobook can and can't be
A Google AI Studio audiobook is a do-it-yourself job: you paste text into Google's speech playground, a Gemini TTS voice reads it, and you download the clip. The voices are among the best available, the free tier costs nothing, and for a short story or a sample chapter that may be all you need.
For a whole book, the work is in the gaps. Gemini TTS voices at most two speakers per request, takes about 8k tokens of input at a time, and knows nothing about chapters. This guide covers the steps, the limits from Google's own documentation as of October 2026, how people stitch around them, and one publishing mistake a popular tutorial makes.
What Google AI Studio's speech generation does
Google AI Studio is the browser playground for the Gemini API, and its speech mode is the quickest way to try Google AI Studio text to speech; Google's docs suggest testing the models there before building anything. They list four TTS models: Gemini 3.8 Flash TTS and 3.8 Flash-Lite TTS (updated September 2026), plus 3.1 Flash TTS Preview and 2.5 Pro Preview TTS. 3.8 Flash covers more than 130 languages and detects the language of your text.
You direct delivery in plain English. A style instruction such as “cheerful and friendly” applies to a turn, and Google documents inline vocal events written in angle brackets, like <laugh>, <sigh> or <whispering>. There are 30 prebuilt voices plus an extended library, and AI Studio also offers voice design and voice replication, which a book-focused tool may not.
Step by step: one chapter in AI Studio
Split the chapter first. Each request is capped at 8,192 input tokens and 16,384 output tokens, so cut the chapter at scene breaks into pieces short enough for one request; if a clip stops early, make the piece shorter. Then, for each piece, choose single-speaker for narration, pick a voice, write a style instruction for the narrator, paste the text, run it and download the audio.
Keep a log of each piece's model, voice and style instruction. Output varies from run to run, so a narrator who sounds warm in scene one can sound brisker in scene three unless every setting matches. Listen to each clip before moving on: a stitched book is much harder to fix than a single clip.
The limits: two speakers, input length, daily caps
Gemini multi speaker TTS is where most fiction hits a wall, and Google's documentation is clear: “Single-request multi-speaker generation … supports up to 2 speakers using prebuilt voices.” For more, it says to synthesize each speaker's turn individually. A scene with a narrator and three characters is already four voices, so most fiction ends up generated turn by turn.
Daily caps are harder to pin down. The TTS models are free of charge on the free tier, but Google's rate-limit page now only says your limits are visible in AI Studio, and community reports in 2026 put the free allowance for some TTS models as low as 10 requests a day. Free-tier inputs and outputs are used to improve Google's products, and paid prices for the 3.8 TTS models double on January 1, 2027.
Stitching chapters and casting more than two characters
To voice a full cast by hand, go through the chapter line by line: decide who speaks each line, generate each speaker's lines with that speaker's voice, then assemble the clips in order in an audio editor. Google notes that each WAV response carries a 44-byte header to strip before joining files in code. Then add chapter markers, even out the loudness and export.
That bookkeeping (who speaks, which voice, what order, where chapters start) is exactly the work multi-voice text to speech tools exist to do. For a novel with dozens of speaking parts, it runs to hundreds or thousands of clips. If you're weighing chatbots too, can ChatGPT make an audiobook covers what they do with a manuscript.
Publishing: why the ACX route in YouTube tutorials breaks the rules
The top video for this search, “Google AI Studio: Create an AI Audiobook and Publish It on Audible!” (July 2025), is a helpful walkthrough of the playground, but its last step tells viewers to publish through ACX. That step doesn't hold up. ACX's audio submission requirements, dated April 15, 2026, say an audiobook “must be narrated by a human unless otherwise authorized” and that “unauthorized use of text-to-speech, AI, or automated recordings in ACX titles is prohibited.”
Audible says on the same page that it is working to accept third-party TTS content and will share timing later, so this may change. Until then, a Gemini-narrated book uploaded to ACX risks rejection or removal. See whether Audible accepts AI narration for the current rules, and publishing on Spotify for a route that takes digital voice narration with a disclosure.
The same Gemini voices with automatic casting and M4B export
If what you want is a Gemini TTS audiobook without the clip management, note that FullCastBook's standard voices run on Google's Gemini TTS, so the sound is the same family you hear in AI Studio. The difference is the work around it. Upload an EPUB, DOCX, PDF with a text layer, TXT or Markdown file, and chapters and speakers are detected for you. Each line is generated with its own speaker's voice, so the two-speaker limit never comes up, and a direction like [dry, unimpressed] goes along as the line's style instruction.
You listen in the browser, regenerate single lines and export an M4B with chapter markers, one MP3 or a ZIP with a file per chapter. Premium ElevenLabs voices are there too if you want a different sound. The free credits on sign-up cover a short sample chapter, and pricing shows what a full book takes.
Sources and dates
Checked on October 21, 2026: Google's speech generation guide (updated October 12, 2026), the Gemini 3.8 Flash TTS model page, the Gemini API pricing page (updated October 19, 2026), rate limits (updated October 15, 2026) and API terms (updated April 28, 2026), and ACX's audio submission requirements (April 15, 2026). Links are under the comparison table. Google updates these models often, so recheck the limits before you start a Google AI Studio audiobook of any length.
