For the complete documentation index, see llms.txt. This page is also available as Markdown.

Multi Speaker Mode

How to configure and generate two-speaker dialogue audio in Qolaba's Speech Generation — voice assignment, the dialogue card system and generating a combined audio output.

Multi-Speaker Mode generates dialogue-based audio using two distinct voice profiles. Instead of a single continuous narration, the output alternates between two speakers — producing a realistic conversation, interview, or dialogue from a structured script.


When to Use It

Multi-Speaker Mode is best for any content that involves a conversation or exchange between two voices:

  • Podcast interviews and co-hosted episodes

  • Q&A and FAQ format audio

  • Story-based scripts with two characters

  • Training simulations and role-play scenarios

  • Conversational product demos

  • Interview-style marketing and promotional content


How the Dialogue Card System Works

Multi-Speaker Mode uses a dialogue card system to structure the conversation. Each card represents one turn in the dialogue and is assigned to either Speaker 1 or Speaker 2. Cards alternate between speakers automatically — the first card is Speaker 1, the second is Speaker 2, the third is Speaker 1 again, and so on.

This structure ensures the voices alternate naturally throughout the conversation, producing a single combined audio file with both speakers in sequence.


Step-by-Step Configuration

Step 1 — Select Multi-Speaker Mode Open the Speech Generation workspace and select Multi-Speaker from the mode selector.

Step 2 — Assign Voices Select a distinct voice for Speaker 1 and a distinct voice for Speaker 2. Choose voices that contrast in tone, energy, or style to make the conversation feel natural and easy to follow.

Step 3 — Set Accent and Style Instructions Accent and style instructions apply across the full conversation — not per speaker. Select the accent that matches the language and region of your script, and add a style instruction that sets the overall tone of the dialogue.

Examples:

Step 4 — Write Your Dialogue

Enter the dialogue for each speaker using the card system:

  1. The first card is pre-assigned to Speaker 1 — enter their opening line or turn

  2. Click Add Dialog — a new card appears assigned to Speaker 2

  3. Enter Speaker 2's response or turn

  4. Click Add Dialog again to add Speaker 1's next turn

  5. Continue until the full dialogue is complete

Multi-language support: Each dialogue card can be written in any language. The audio for that card will be generated in the language of the text entered — allowing multilingual dialogue if needed.

Script writing tips for dialogue:

  • Write the way people actually speak — contractions, natural pauses, and conversational phrasing produce more realistic audio

  • Use punctuation deliberately — commas for brief pauses, full stops for complete breaks, question marks for rising intonation

  • Avoid overly long turns per card — break lengthy speaker contributions across multiple cards if needed

  • Add emotional cues through the style prompt rather than trying to describe them in the script text itself

Step 5 — Select a Model

Model
Best For

Flash TTS

Draft generation, testing dialogue flow and voice pairing

Pro TTS

Final production output, client delivery, publishing

Step 6 — Generate Click Generate. The system processes all dialogue cards in sequence and produces a single combined audio file with both voices alternating as configured.


Reviewing and Using Your Output

Once generated, you can:

  • Play the combined audio to review the full dialogue

  • Download the file to your device

  • Share via a shareable link

  • Regenerate by adjusting individual dialogue cards, voice assignments, or style instructions if the output needs refinement

All generated audio is automatically saved to the Audio History panel and to Dashboard → History for future access. See Managing Generated Audio → for full details.


Resetting

The Reset option clears all dialogue cards, voice assignments, style instructions, and model selection. Use this when starting a new dialogue from scratch.

Last updated