Gemini TTS

PODCAST / FROM YOUR WRITTEN SCRIPT

Create Two-Host Podcast Audio with Gemini TTS

Bring a written podcast script into a conversation. Choose two voices, assign each turn, and use Gemini TTS to generate a recording you can review and download as WAV. Start with an editable opening, interview, or explanation below; your script stays yours to change.

Flash · English

A demonstration generated for this site · 2026-10-08

Read the full transcript

Welcome to Small Changes, the show where we turn one practical question into a useful experiment. Today we are asking why a short first draft can help a project move forward. Because a smaller draft gives you something concrete to show. Instead of explaining every possible feature, you can ask whether one part solves the problem. So our challenge is simple: choose one idea, write one example, and share it with someone who needs it. Exactly. By the end of this conversation, you will have a first step you can actually try.

Model

Script

1 characters / 5,000

Two speakers

Speaker 1

Voice sample uses neutral settings.

Speaker 2

Voice sample uses neutral settings.

Advanced controls
Estimated credits1.2

Samples use no credits. Generating does.

Start a Gemini TTS podcast with a focused script

A useful podcast segment gives listeners one question to follow. Write an opening that names the topic, let the second host add a concrete example, and finish with a clear takeaway. This page starts from your written script. It does not turn a topic, uploaded document, or website into an automatically researched episode. If you have notes rather than finished lines, turn those notes into a conversation before bringing them into the studio.

Keep the first test short enough to review carefully. A few turns reveal whether the roles are distinct, the names are pronounced correctly, and the delivery fits the subject. The studio accepts up to 5,000 characters in one request, counting the combined dialogue text and vocal tags. A longer programme can be prepared as separate segments for your editor, but adjacent recordings still need a listening check for changes in voice or pacing.

A two-host podcast opening

Introduce the topic and let the second host explain what listeners will learn.

Speaker 1: Welcome to Small Changes, the show where we turn one practical question into a useful experiment. Today we are asking why a short first draft can help a project move forward. Speaker 2: Because a smaller draft gives you something concrete to show. Instead of explaining every possible feature, you can ask whether one part solves the problem. Speaker 1: So our challenge is simple: choose one idea, write one example, and share it with someone who needs it. Speaker 2: Exactly. By the end of this conversation, you will have a first step you can actually try.

Scene direction: A friendly podcast opening. The host is curious; the co-host gives clear, measured replies.

Give each host a reason to speak

Two hosts are easier to follow when their roles differ. One might ask the question while the other explains a practical detail. Give each turn a speaker number, choose a different preset voice for each role, and keep that assignment consistent throughout the script. Repeating the same claim in two voices adds length without helping the listener. Instead, let a reply explain, challenge, clarify, or connect what the first host has said.

Listen to the voices together before approving a take. An energetic host and a steady guest can work well, but the labels in the voice library are only starting points. A neutral preview does not guarantee the exact delivery of your script. Keep optional controls simple for the first generation, then change one setting at a time. Voice descriptions apply to each speaker; scene direction describes the situation they share.

Build an interview around one useful answer

A short interview works best when the question leaves room for a specific answer. Ask what changed, what the guest tried, or what a listener could do next. Follow up on the answer rather than jumping to an unrelated question. The template below gives you an editable structure; replace the subject and details with your own verified material. Neither a confident voice nor a polished exchange checks the factual accuracy of the words.

Use brief stage direction to describe the relationship between the speakers. A curious host and a thoughtful guest need less instruction than a long list of competing emotions. Read both parts aloud before generating. Remove written shortcuts, split sentences that are difficult to say, and explain unfamiliar terms in the script. A natural conversation still benefits from clear editing, even when you want it to sound informal.

A host and guest interview

Use a focused question, a concrete answer and a follow-up.

Speaker 1: What did you change when your first plan became too complicated? Speaker 2: I chose one person who needed help and asked them to describe the last time the problem happened. Their example was much clearer than my list of ideas. Speaker 1: What did you make after that? Speaker 2: A short version that handled that one situation. We tested it together, wrote down what was confusing, and improved the explanation before adding anything else.

Scene direction: A thoughtful interview between a curious host and a relaxed guest.

Make room for pauses and reactions

Punctuation and sentence length provide a good first guide to pacing. Put a supported <short pause>, <sigh>, or <laugh> cue at the place where it belongs in the speaker’s line. Use a pause to make an idea easier to hear, rather than adding one after every sentence. Tags are submitted as part of the script and contribute to its length. Listen to the result to decide whether the instruction helped.

For an explanation, let the second speaker ask for a distinction or a concrete example. This gives the first speaker a reason to slow down without repeating the opening. The knowledge template below demonstrates that pattern. Applying a template only fills editable text and settings; it does not generate audio or spend generation credits. Review the selected model and the live estimate after editing, then submit only when the passage is ready.

Explain one idea together

Let a question lead to an explanation and one practical example.

Speaker 1: What is the difference between a goal and a next step? Speaker 2: A goal describes where you want to arrive. A next step describes one action you can take now. Speaker 1: Can you give a simple example? Speaker 2: Learning to cook is a goal. Choosing one recipe and preparing the ingredients is a next step. You can check whether that action is complete, then decide what follows.

Scene direction: Two podcast hosts explain a practical idea clearly, without rushing.

Review and download your podcast recording

When the task completes, listen to the full WAV recording in the studio. Check the opening, every speaker change, names and numbers, and the closing sentence. If one section is unclear, revise that short passage and generate a new take rather than assuming extra direction will fix everything. Completed recordings also appear in your library. Download the version you want to keep before moving on to another variation.

The two-speaker request produces one combined audio file. It does not deliver separate tracks for the hosts, automatically add music, publish an RSS feed, or synchronise a video. Use your audio editor for those production steps. You can place the WAV clip on a timeline, trim the opening and ending, and review levels alongside other material. Any MP3 conversion or final episode assembly belongs to that later editing workflow.

Choose a model and review the credits

Both available models support the single-voice and two-speaker controls in this studio. Start with Flash when you want to explore expressive delivery; compare Flash-Lite at its lower configured credit rate. The model comparison page provides controlled scripts and application credit examples. Keep the same passage and voices when testing both models, so differences in the writing do not get mistaken for differences between models.

Credits are based on started blocks of 100 submitted characters. A 101-character script starts a second block; all dialogue turns contribute to the total. The estimate updates as you edit and switch models. New accounts receive 20 starter credits, valid for 30 days from registration. Generation requires authentication and sufficient available credits. The pricing page explains this application’s rates; they are distinct from Google’s direct API pricing.

Questions before you start

Can Gemini TTS write the podcast for me?

This application generates speech from your written script. Use the editable examples to structure your conversation; topic research, automatic writing and document import are not part of this workflow.

How many hosts can I use?

Up to two speakers in one request. Assign every turn to speaker one or two and select a voice for each. The result is one combined WAV recording.

Do the templates use credits?

Copying or applying a template does not submit a generation. Only generate after reviewing the text, selected model and credit estimate.

Can I make a full episode with music?

Prepare longer episodes as reviewed segments within the 5,000-character request limit, then assemble them in an editor. Automatic music, separate speaker tracks and publishing are not provided here.