Dramabox brings actor-style emotional delivery to open-weight text-to-speech
ResembleAI released Dramabox, an open-weight text-to-speech model built on LTX-2 that generates audio with emotional inflection, dramatic pauses, and theatrical delivery—no API keys or usage limits required.
ResembleAI released Dramabox, an open-weight text-to-speech model that generates audio with emotional range and theatrical delivery. Built on LTX-2, the model adds intonation, dramatic pauses, and accent variation to synthetic speech—capabilities typically reserved for closed commercial APIs like ElevenLabs or Play.ht. The weights, inference code, and a live HuggingFace Space demo are all public, with no usage caps or API keys required.
Dramabox runs locally without API calls. Users can control pacing, add dramatic pauses, and shift emotional tone mid-sentence—features that typically require expensive studio voice actors or closed-source tools with per-minute billing. The GitHub repository includes setup instructions and example scripts for generating emotionally varied speech from text prompts.