DramaBox: Direct a Voice Performance
Turn a scene-style prompt into a performed audio clip with DramaBox, Resemble AI's expressive open-source TTS model. Describe your speaker and delivery, optionally clone a voice, and hit run.
dramabox
expressive speech
opensource
resemble ai
text to speech
0
58
Direct a Voice Performance
Write a prompt describing your speaker and how they deliver each line, with the spoken words in double quotes and the acting notes outside them. Add an optional voice clip if you want to clone a specific voice. Hit run and get a finished audio performance back.
Model
DramaBox by Resemble AI. An open-source expressive text-to-speech model that turns scene-style prompts into full vocal performances, with emotional arcs, natural pacing, breaths, and laughs. It can also clone a voice from a short reference clip.
HOW IT WORKS
Step 1. Write your scene prompt
Describe who is speaking, the emotion, and the delivery. Put the spoken lines in double quotes and the stage directions outside them. Example: A tired detective mutters, "I already told you everything I know," then sighs.
Works great with: dialogue · narration · character voices
Step 2. Add a voice reference (optional)
Upload 10 or more seconds of clean audio to clone a specific voice. Skip it and the model invents a voice to match your speaker description.
Step 3. Hit run and download
The model performs your scene and returns a 48kHz stereo audio file. Preview it in the workflow, then download.
Ready for: video editors · game engines · podcast tools
First time? Leave every setting as-is. The defaults are the right starting point for almost everyone.
RECOMMENDED SETTINGS
Quick-start guide. Find the goal that matches yours and copy the settings.
Standard performance (most people). 30 steps · CFG 2.5 · duration 0 · random seed. The right starting point for almost everyone.
Want the model to set the length. Leave duration at 0 so it times the audio to your prompt instead of forcing a fixed runtime.
Want to lock a take you liked. Set a specific seed instead of random, so the same prompt reproduces the same performance.
Want tighter control over the delivery. Raise the CFG scale so the output follows your scene direction more closely.
The pacing feels rushed or too slow. Adjust speed away from 1, or rewrite the pacing cues in your prompt.
The cloned voice sounds noisy. Use a cleaner reference clip. Background noise in the sample bleeds into the result.
Prompt: Put spoken lines inside double quotes and stage directions outside them, where they steer the performance without being read aloud. Write phonetic sounds like laughs as single words inside the quotes, such as "Hahaha". A nervous student whispers, "I think I broke something," then a long pause lands better than say this sadly: I broke something.
LEARN
📹 Videos
ComfyUI 101 Free Course ft. Sebastian Kamph
Floyo 101 for Team Collaboration
✨ Quick links
USE CASES
🎧 Audiobook & Narration
Narrate a passage with the emotional arc the scene calls for, instead of a flat read.
🎮 Game & Animation Dialogue
Voice characters from a written prompt and clone a reference to keep each one consistent across lines.
🎬 Video & Explainer Voiceover
Generate a voice track for a cut, with the pacing and tone matched to the moment.
🎙️ Podcast & Character Voices
Build distinct speakers for fiction, skits, or intros without recording every take yourself.
WHAT WORKS BEST / WHAT TO AVOID
✅ Works great
Scene prompts with clear emotion and delivery notes
Spoken lines wrapped in double quotes
Clean 10-second-plus reference clips for cloning
English dialogue and narration
⚠️ May produce softer results
Plain text with no direction or emotion
Noisy or muffled reference audio
Long stretches of dialogue with no pacing cues
Rare accents or speech patterns underrepresented in training data
FAQ
What is DramaBox?
DramaBox is an open-source expressive text-to-speech model from Resemble AI. It turns a single prompt into a full vocal performance covering identity, emotion, laughs, sighs, breaths, and transitions. It is an IC-LoRA fine-tune of the LTX-2.3 audio model built by Lightricks, and it can clone a voice from a short reference clip. Resemble AI
How do scene prompts work in DramaBox?
The prompt acts like a director's script. Dialogue goes inside double quotes and is spoken literally, while stage directions go outside the quotes and function as performance cues that are never spoken. Phonetic sounds such as laughter go inside the quotes as single words, like "Hahaha" rather than "ha ha ha". This lets you steer emotion and pacing in plain language. Resemble AI
Do I need a voice sample to use it?
No. If you provide 10 or more seconds of reference audio, DramaBox applies that timbre to your prompt's performance. Skip the reference and the model invents a voice to match the description. A clean reference clip clones better, since any background noise in the sample carries into the output. Resemble AI
Does DramaBox add a watermark to the audio?
Yes. Every clip gets a Resemble Perth watermark that the company says survives MP3 compression and routine edits at near-100% detection accuracy. It marks the audio as AI-generated without changing how it sounds. aiHola
What languages does DramaBox support?
DramaBox is built for expressive English speech and is strongest there. Accent and dialect are captured, but accent accuracy degrades for highly unusual or underrepresented speech patterns not well represented in training data. MindStudio
Can I use DramaBox audio commercially?
DramaBox is open-weight and released by Resemble AI, built on top of the LTX-2.3 audio branch under the LTX-2 Community License. Commercial use is generally allowed, but review the current license terms, and note that outputs carry the embedded Perth watermark.
How to run DramaBox online?
You can run DramaBox online through Floyo. No installation, no setup, no API key to wire up. Open the workflow in your browser, write your scene prompt, and hit run. Free to try.
WHY FLOYO?
Floyo is the only platform with team collaboration for ComfyUI in the browser. You run workflows with no install. You share run history, assets, and models across your team. You pay only when you generate. Floyo supports open-source and closed-source models.
A creator runs a voice take and likes the result. A teammate opens that exact run from shared history and keeps going. No file handoffs. No version confusion.
For studios and enterprise teams, Floyo adds private workspaces, pooled resources, and a team usage dashboard. Other ComfyUI cloud tools run for one person at a time. Floyo runs for the whole team, with transparent per-generation costs.
Ready to try it?
Write your scene, add an optional voice clip, and run it. The settings are already set.
Questions? Watch the free course or check the FAQ above.
Read more
