OpenMOSS / MOSS-TTSD
MOSS-TTSD
MOSS-TTSD is a production long-form dialogue model for expressive multi-speaker conversational audio at scale. It supports long-duration continuity, turn-taking control, and zero-shot voice cloning from short references for podcasts, audiobooks, commentary, dubbing, and entertainment dialogue.
# MOSS-TTSD (Multi-Speaker Dialogue TTS)
Long-form spoken dialogue generation model with multi-party conversational speech synthesis. Part of the MOSS-TTS Family — an open-source speech and sound generation model family from MOSI.AI and the OpenMOSS team. MOSS-TTSD supports 1-5 speakers with flexible control, handles natural turn-taking, and features zero-shot voice cloning with only 3-10 seconds of reference audio. Supports Chinese, English, Japanese, and European languages (auto-detected from input text).
dialogue (string, required) — Dialogue text with speaker tags. Use [S1], [S2] etc. to indicate speaker turns. Example: "[S1] Hey, did you finish the report? [S2] Almost done, just need to review the numbers."
s1Audio (file, optional) — Speaker 1 reference audio for voice cloning. Must be provided together with Speaker 1 Text. Supports .wav, .mp3, .m4a and .webm formats.
s1Text (string, optional) — Speaker 1 prompt transcript for voice cloning. Must be provided together with Speaker 1 Audio. The [S1] tag is added automatically if missing.
s2Audio (file, optional) — Speaker 2 reference audio for voice cloning. Must be provided together with Speaker 2 Text. Supports .wav, .mp3, .m4a and .webm formats.
s2Text (string, optional) — Speaker 2 prompt transcript for voice cloning. Must be provided together with Speaker 2 Audio. The [S2] tag is added automatically if missing.
s3Audio (file, optional) — Speaker 3 reference audio for voice cloning. Must be provided together with Speaker 3 Text. Supports .wav, .mp3, .m4a and .webm formats.
s3Text (string, optional) — Speaker 3 prompt transcript for voice cloning. Must be provided together with Speaker 3 Audio. The [S3] tag is added automatically if missing.
s4Audio (file, optional) — Speaker 4 reference audio for voice cloning. Must be provided together with Speaker 4 Text. Supports .wav, .mp3, .m4a and .webm formats.
s4Text (string, optional) — Speaker 4 prompt transcript for voice cloning. Must be provided together with Speaker 4 Audio. The [S4] tag is added automatically if missing.
s5Audio (file, optional) — Speaker 5 reference audio for voice cloning. Must be provided together with Speaker 5 Text. Supports .wav, .mp3, .m4a and .webm formats.
s5Text (string, optional) — Speaker 5 prompt transcript for voice cloning. Must be provided together with Speaker 5 Audio. The [S5] tag is added automatically if missing.
textNormalize (checkbox, optional) — Normalize dialogue text before synthesis. Removes special characters and fixes punctuation for cleaner output.
API quick start
Run MOSS-TTSD with a single API call.
{
"dialogue": "[S1] Hey, did you finish the report? [S2]…",
"inputAudio1": "https://your-cdn.com/input.mp3",
"inputText1": "...",
"inputAudio2": "https://your-cdn.com/input.mp3"
}