In short: Switch on Two speakers in the bottom-left corner of the Text to Speech box. Start each line with
A: or B:, tap A and pick a voice, tap B and pick a second voice, type the security code and press Generate Audio. You get one MP3 with both voices, up to 2,000 characters per conversion.Make your first dialogue in 2 minutes
- Switch it onIn the Text to Speech box, press the small Two speakers switch in the bottom-left corner of the text area. A bar opens with Speaker A and Speaker B. On this page it is already on.
- Write two linesType
A: Hi, welcome to the show!, press Enter, thenB: Thanks for having me.Each new line that starts with a speaker label is a new turn. - Choose voice ATap the A box in the bar, then tap any voice in the voice list. Press ▶ on a voice first to preview it.
- Choose voice BTap the B box and pick a second voice. Speechma already suggests a matching partner (same language and country, other gender), so often you only change it if you want to.
- GenerateType the 5-digit security code once and press Generate Audio. The finished MP3 appears in your results list, ready to play or download.
Every control in the Two Speakers bar
| Control | What it does |
|---|---|
| Two speakers switch | Turns dialogue mode on or off. Speechma remembers your choice and your two voices on this device. |
| A and B boxes | Show each speaker’s voice. Tap one, then tap a voice in the list to give it to that speaker. The chosen voices carry a small A or B tag in the list. |
| ⇄ (swap) | Swaps the two voices, so A speaks B’s lines and the other way round. |
| + A / + B | Adds a new line for that speaker at the end of your script, ready to type. |
| Pause | The silence between speakers: Short (0.15 s), Natural (0.35 s) or Long (0.7 s). Long suits language lessons. |
| Voice Effects | Pitch and speed apply to both speakers. |
How to write speaker labels
Speechma reads the start of each line. All of these work:
| You write | Speechma understands |
|---|---|
A: and B: | Speaker A and Speaker B |
1: and 2:, Speaker 1: and Speaker 2: | Speaker A and Speaker B |
Any two names: Host: / Guest:, Mia: / Leo: | The first name you use is A, the second is B |
Names in brackets: [Teacher] / [Student] | Same as names with a colon |
| A line with no label | Continues the speaker of the line above |
| No labels at all | Lines take turns: A, B, A, B |
Times like 10:30 and web addresses like https:// are read as normal text, not as speaker names.
Limits
| What | Limit |
|---|---|
| Speech per conversion | 2,000 characters (labels like “A:” do not count) |
| When an AI voice is used | Up to 1,000 characters for the AI voice lines |
| Turns per dialogue | Up to 40 (lines in a row from the same speaker count as one turn) |
| Speakers | 2. For a third voice, make its lines as a separate clip |
| Security code | One code for the whole dialogue |
Longer script? Split it into parts and join the MP3s in any editor, or use SpeechmaPRO.com for 10,000 characters at once.
Tips for natural-sounding conversations
- Contrast the voices. A male and a female voice, or two clearly different tones, help listeners follow who is talking.
- Same country, same language. Two American voices sound like they are in the same room; mixing countries works best for interviews across borders.
- Keep turns short. One or two sentences per turn sounds like a real conversation. Long speeches sound read.
- Add small reactions. “Right.” “Wait, really?” “Makes sense.” They make dialogue feel alive.
- Punctuation is direction. Commas give short pauses, periods longer ones, question marks lift the voice.
- Fix one line, keep the rest. Change only the line you do not like and generate again: lines that did not change are reused instantly.
Fix common problems
| You see | Do this |
|---|---|
| “This script has more than two speakers” | A line starts with a word and a colon, like “Note:”. Remove the colon or start that line with A: or B: |
| “A dialogue needs lines for both speakers” | Add at least one line that starts with B: (or your second name) |
| “Choose a voice for Speaker B” | Tap the B box, then tap a voice in the list |
| The wrong voice speaks a line | Check that the line starts with the right label. A line with no label continues the speaker above it |
| “Type the security code again” | The code is valid for 10 minutes. Type the new code and press Generate again |
| The text is too long | Split the dialogue into two parts of up to 2,000 characters each |
Use voices responsibly
Dialogue audio is yours to use in videos, podcasts, courses, ads and apps, including commercial projects. Never make a dialogue that pretends to be a real, named person, and tell your audience when content uses AI voices where that matters. See our AI voice ethics page.
Frequently asked questions
Where is the Two speakers switch?
In Text to Speech, inside the text box, in the bottom-left corner next to the character count.
Can both speakers use the same voice?
Yes, but listeners will not hear who is talking. Two different voices sound much better.
Does Two Speakers work with AI voices?
Yes. Any voice in the list works, standard or AI, and you can mix them. AI voice lines can have up to 1,000 characters in total.
Why is the result one file and not two?
Speechma joins every line in the right order, with the same loudness and natural pauses, so you get a finished conversation you can use right away.
Is Two Speakers free?
Yes. It is part of the free Text to Speech tool. No account is needed.
Can I make a podcast with it?
Yes. See the AI podcast generator guide and the dialogue script guide for templates.
The Speechma newsletter
Make better voiceovers, faster
- New voices and features the day they launch
- Script templates and tricks that make AI voices sound human
- Creator stories, community picks and early access
