Cookie Preferences
We use cookies to personalize content and analyze traffic. See our Cookie Policy.
No audio yet. Submit the form to generate audio.
Explore different use cases and parameter configurations
Complete guide to using elevenlabs/text-to-dialogue-v3
Build natural, emotionally expressive dialogue experiences across 70+ languages with fine-grained control, multi-speaker generation, and flexible deployment through APIPASS.
The Eleven V3 (Alpha) API is ElevenLabs' most expressive Text to Speech model to date, engineered to sound less like mechanical synthesis and more like a real, directed performance. It surpasses traditional AI voice generation by understanding emotional nuance, pacing, and context, allowing generated voices to whisper, laugh, sigh, and react naturally. Powered by APIPASS, this API brings advanced multi-speaker Text to Dialogue and inline audio tags directly into your production workflows, enabling you to build immersive, emotionally rich audio storytelling and interactive applications.
Every request must include dialogue as a non-empty array. Each item must be an object containing both text and voice.
The adapter only accepts stability values of 0, 0.5, or 1. If the field is omitted, the adapter injects the default value 0.5 before forwarding the task upstream.
You can send language_code to guide the provider. If you omit it, the adapter leaves the field out and allows automatic language detection.
Creators can use inline tags within the text to apply fine-grained control over tone, emotion, and non-verbal reactions, making the speech feel intentional rather than artificially generated.
The API natively supports multi-speaker Text to Dialogue, enabling flowing conversations with realistic turn-taking, pacing, and interruptions without the need to manually stitch separate voice tracks together.
The model maintains emotional delivery, nuance, and prosody across more than 70 languages, making it perfectly suited for international audiences and multilingual products.
Eleven V3 interprets context and intent from scripts exceptionally well, resulting in improved stress, cadence, and emotional continuity across longer spoken passages.
Developers can choose from a variety of audio formats to suit their application needs, including MP3, PCM (S16LE), μ-law, A-law, and Opus.
Instead of locking users into rigid subscription plans, the platform utilizes a flexible credit-based system, allowing teams to seamlessly scale usage and control spending based on real demand.
The integration follows the standard ApiPass asynchronous jobs flow used by Kie-backed generation models.
Create an account on apipass.dev and generate your unique API Key to unlock access and authenticate your requests to the Eleven V3 API.
Utilize the apipass.dev playground to experiment with expressive generation, test out various audio tags, and preview multi-speaker dialogue behavior before integrating it into your live production environment.
Define your request structures within your application by preparing the dialogue text inputs, setting up the speaker structure, and selecting the appropriate output formats for your specific use case.
Deploy the API directly into your service, application, or internal workflow to power voice features across your creative or interactive platforms.
As your product grows, you can easily scale your Eleven V3 API usage on apipass.dev, utilizing the same technical foundation for both simple voice generation and advanced dialogue experiences.
These suggestions line up with the adapter guardrails and the upstream payload shape.
To ensure the most reliable generation and preserve expressive quality, keep your text requests at or below 2,000 characters and divide long scripts into logical segments like scenes or speaker turns.
Place audio tags (like [whispers] or [laughs]) intentionally at key emotional beats rather than using them continuously, as overusing tags can make the dialogue feel overly stylized and less consistent.
nserting ellipses is an effective way to influence pacing, introducing subtle trailing thoughts, pauses, or hesitation, which is crucial for conversational timing.
Dashes help indicate abrupt changes in direction or mid-sentence interruptions, producing a highly realistic conversational flow and dynamic rhythm.
Perfect for long-form narrative content, allowing creators to produce audiobooks and stories that carry distinct character emotion and pacing closer to an actual performance.
Easily generate podcast-style episodes or interview simulations with realistic multi-speaker timing and emotional continuity.
Bring video game characters and NPCs to life with believable personalities, branching conversations, and narrative-driven interactions that significantly enhance player immersion.
Align expressive Text to Speech voiceovers seamlessly with visual storytelling for AI avatars, video narration, and creative media workflows where delivery deeply defines the user experience.
The document uses elevenlabs/text-to-dialogue-v3 because that is the exact public model identifier recognized by the Kie adapter for this feature.
The readme and API sections describe the same rules implemented in code: dialogue is required, each item needs text and voice, total text must stay within 5000 characters, and stability only accepts 0, 0.5, or 1.
The examples return resultUrls because the Kie adapter normalizes completed task output into the common ApiPass resultJson structure used across asynchronous generation models.
All APIs require authentication via Bearer Token.
Authorization: Bearer
Create a new ElevenLabs Dialogue V3 audio generation task through the ApiPass unified jobs API
ApiPass keeps the public request envelope stable across asynchronous Kie-backed models. Dialogue generation parameters are passed inside the input object.
1{
2 "model": "elevenlabs/text-to-dialogue-v3",
3 "callBackUrl": "string (optional)",
4 "input": {
5 "dialogue": [
6 {
7 "text": "string",
8 "voice": "string"
9 }
10 ],
11 "stability": "0 | 0.5 | 1 (optional)",
12 "language_code": "string (optional)"
13 },
14 "channel": "auto"
15}modelRequiredstringThe public ApiPass model name for ElevenLabs dialogue generation.
"elevenlabs/text-to-dialogue-v3"
callBackUrlOptionalstringOptional callback URL for task completion notifications.
"https://your-domain.com/api/callback"
channelOptionalstringYou may specify the corresponding provider within APIPASS via the channel parameter; these providers handle the actual image and video generation tasks. APIPASS currently offers three provider options:
The default value for the channel parameter is auto. When enabled, APIPASS automatically allocates tasks across available providers based on real-time pricing and stability metrics to balance minimal cost and reliable performance. Retain the default auto value unless you have custom routing requirements.
Available options:
auto
The input object contains the dialogue-generation fields accepted by ApiPass. The adapter validates required dialogue items and injects the default stability value when omitted.
input.dialogueRequiredarray<object>Dialogue array. Each item must contain text and voice, and the total combined text length across all items must not exceed 5000 characters.
[{"text":"Hello and welcome.","voice":"BIvP0GN1cAtSRTxNHnWS"},{"text":"Thanks, let's begin.","voice":"aMSt68OGf4xUZAnLpTU8"}]input.stabilityOptionalnumberOptional stability control. Allowed values are 0, 0.5, and 1. The adapter defaults to 0.5 when this field is omitted.
Available options:
0.5
input.language_codeOptionalstringOptional language code forwarded upstream when provided.
"eng"
1curl -X POST "https://api.apipass.dev/api/v1/jobs/createTask" \
2 -H "Content-Type: application/json" \
3 -H "Authorization: Bearer YOUR_API_KEY" \
4 -d '{
5 "model": "elevenlabs/text-to-dialogue-v3",
6 "callBackUrl": "https://your-domain.com/api/callback",
7 "input": {
8 "dialogue": [
9 {
10 "text": "Thanks for joining the launch review today. I want us to keep the tone warm and natural.",
11 "voice": "BIvP0GN1cAtSRTxNHnWS"
12 },
13 {
14 "text": "Absolutely. I will keep the pacing calm and conversational so the exchange feels like a real discussion.",
15 "voice": "aMSt68OGf4xUZAnLpTU8"
16 }
17 ],
18 "stability": 0.5,
19 "language_code": "eng"
20 }
21 }'1{
2 "code": 200,
3 "message": "success",
4 "data": {
5 "taskId": "task_elevenlabs_dialogue_1768468200000"
6 }
7}codeStatus code. 200 indicates the task was created successfully.
messageResponse message returned by ApiPass.
data.taskIdTask identifier used for later status checks.
