Nodaro DocsText to Dialogue
Voice a whole conversation in one audio file, with a voice for each line and audio tags for emotion. Runs ElevenLabs Dialogue v3 or Dialogue v4.
The Text to Dialogue node voices a whole conversation in one audio file. You write the lines, give each line a voice, and ElevenLabs Dialogue v3 or ElevenLabs Dialogue v4 speaks them in order, as one natural exchange. Use it for interviews, podcast-style conversations and dialogue scenes, instead of generating and stitching separate speech clips.
When to use it
- A podcast-style conversation between two or more speakers.
- An interview with a host voice and a guest voice.
- A dialogue track for an animated video or an explainer.
- A dialogue scene in an audiobook, with one voice per character.
- A quick audio prototype of a conversation.
For one voice reading a text, use Text to Speech.
Quick start
Add the node
Press Tab on the canvas and choose Audio › Speech & Voiceover › Text to Dialogue.
Write the lines
Open the settings panel. Under Dialogue Lines, choose a voice for the first line and type its text. Click Add Line for each new line of the conversation.
Or fill the lines from a script
Connect the Dialogue output of a Generate Script node to the Prompt input. In the settings panel, click the button that fills the lines, such as Fill 12 Lines from Script. Then choose a voice for each speaker.
Run it
Click Run on the node. The node returns one audio file that contains every line in order.
Inputs
| Input | Accepts | What it does |
|---|---|---|
| Prompt | Text nodes, such as Generate Script | When the Dialogue output of a Generate Script node is connected, the settings panel offers a button that fills the lines from the script. The lines in the panel are what the node speaks. |
| Output | What it carries |
|---|---|
| Audio | One continuous audio file with every line spoken in order. |
| Transcript | The timings of the dialogue: each word with its start time, end time and speaker, and one segment for each line. The speaker is the voice of that line. |
Wire the Transcript output into the Transcript input of Add Captions. Add Captions then uses the dialogue's own words and timings instead of transcribing the audio again. Audio tags in the lines never become caption words.
Settings
| Setting | What it does |
|---|---|
| Model | ElevenLabs Dialogue v3 (the default) or ElevenLabs Dialogue v4. See Models. |
| Dialogue Lines | The lines of the conversation, in order. Each line has its own voice and text. Type [ or / in a line to insert an audio tag. Empty lines are skipped. |
| Add Line | Adds a line at the end. The trash button removes a line. |
| Fill … Lines from Script | Appears when a Generate Script node's Dialogue output is connected. Replaces the lines with one line per line of the script. Each line keeps the speaker's name and starts with the default voice, Sarah. |
| Stability | Lower is more dramatic, higher is more even. The default is 0.5. On Dialogue v3, choose Most Variable (0), Balanced (0.5) or Most Stable (1.0). On Dialogue v4, a slider takes any value from 0 to 1. When you switch to v3, a value between the steps moves to the nearest step. |
| Similarity | Dialogue v4 only. From 0 to 1, how closely each line keeps the character of its voice. The default is 0.75. |
| Language | Auto-detect (the default), or one language. |
| Seed | A whole number from 0 to 4,294,967,295. The same seed, lines and settings reproduce the same audio. Leave it empty for a random take. |
| Text Normalization | Auto (the default), On or Off. When on, numbers, dates and abbreviations are spelled out so that they sound natural. |
The panel counts the voices and characters as you write, for example 2/10 voices and 840/5000. The character count shows the limit of the chosen model.
Limits
| Limit | Value |
|---|---|
| Text across all lines | 5,000 characters on Dialogue v3, and 10,000 characters on Dialogue v4. Up to 2,000 characters gives the best quality. |
| Different voices per run | 10. Using a voice on several lines counts once. |
A dialogue needs at least one line with text. A run with no text in any line is refused before it starts. A script over the chosen model's limit is also refused before any credits are charged.
Voices and audio tags
Every line can use any voice: premade voices, Voice Library voices and your own voices, in any mix. There is no separate list of dialogue voices.
On both models, lines understand the same audio tags as ElevenLabs v3 and v4 in Text to Speech, such as [laughs], [whispers] and [sighs].
Models
Text to Dialogue runs one of two models. Both speak the same 46 languages and understand audio tags.
- ElevenLabs Dialogue v3 is the default. It is expressive, and its Stability has three steps.
- ElevenLabs Dialogue v4 is the newest model. It takes any Stability from 0 to 1, adds Similarity, and accepts scripts of up to 10,000 characters.
| Model | Maker | Modes | Credits | Details |
|---|---|---|---|---|
| ElevenLabs Dialogue v3 | ElevenLabs | Multi-speaker dialogue | from 5 | Multi-speaker dialogue via the direct ElevenLabs API — give it a script, it voices each role (any voice: premade, library, or cloned). |
| ElevenLabs Dialogue v4 | ElevenLabs | Multi-speaker dialogue | from 5 | Multi-speaker dialogue on ElevenLabs v4 — [audio tags], stability and similarity control. Direct API. |
Credits
The price follows the length of the script. Every started block of 100 characters across all lines costs 5 credits on Dialogue v3 and 5 credits on Dialogue v4. A run is billed as at least 8 blocks, so every script of up to 800 characters costs the same. The Transcript output costs nothing extra. The node shows the exact price before it runs.
Tips
- Give each speaker a clearly different voice. A listener follows the conversation more easily. Preview the voices in the voice browser before you assign them.
- Keep lines conversational. Several short lines sound more natural than one long paragraph in one line.
- Start with Balanced stability. Lower it for a dramatic or emotional scene, and raise it for a formal, narration-like read. On Dialogue v4, you can choose any value in between.
- Set the language for mixed-language talk. Auto-detect works well for one language. When the speakers switch languages, choose the main language explicitly.
- Join long conversations. Split a long conversation across several nodes and join the parts end to end with Combine Audio.
From the API
The same generation is available as the MCP tool generate_dialogue. See the MCP tools and Voice and media.
POST /v1/text-to-dialogue takes dialogue ([{ text, voice }], in speaking order) and the optional provider, stability, similarityBoost, languageCode, seed and applyTextNormalization. provider is elevenlabs-dialogue (Dialogue v3, the default) or elevenlabs-dialogue-v4. The MCP tool calls the same choice model, and the CLI calls it --model on nodaro voice dialogue. On Dialogue v3, a stability other than 0, 0.5 or 1 is refused. The finished job's output_data.transcript holds the same timings as the Transcript output.
Frequently asked questions
How many voices can one dialogue have?
Up to 10 different voices per run. Using the same voice on several lines counts once, so a two-person conversation of 40 lines uses only 2 of the 10.
How long can a dialogue be?
Up to 5,000 characters across all lines on ElevenLabs Dialogue v3, and up to 10,000 characters on ElevenLabs Dialogue v4. Up to 2,000 characters gives the best quality. Split a longer conversation across several Text to Dialogue nodes and join them with Combine Audio.
How much does a dialogue cost?
The price follows the length of the script. Every started block of 100 characters across all lines costs 5 credits, on either model, and a run is billed as at least 8 blocks. The node shows the exact price before it runs.
Can I get each line as a separate audio file?
No. Text to Dialogue returns one continuous file with every line in order, and its Transcript output tells you where each line starts and ends. For separate clips, use one Text to Speech node per line.
Can I turn a written script into dialogue lines automatically?
Yes. Connect the Dialogue output of a Generate Script node to the node's Prompt input. In the settings panel, click the button that fills the lines, such as Fill 12 Lines from Script, then choose a voice for each speaker.
How do I get the same take again?
Set a Seed. The same seed, the same lines and the same settings reproduce the same audio. Leave Seed empty for a new take on every run.
Related
Text to Speech
Generate Script
Combine Audio
ElevenLabs Dialogue v3
ElevenLabs Dialogue v4
Last updated on
Text to Speech
Turn text into natural speech with ElevenLabs v4, v4 Turbo, v3, Turbo v2.5 or Multilingual v2. Choose a voice, add emotion with audio tags, speak 46 languages.
Voice Changer
Replace the voice in a recording or a talking video with another voice, and keep the original emotion, pacing and timing. Speech-to-speech with ElevenLabs.