dots.tts.edit
dots.tts.edit is a continuous autoregressive model for precise, instruction-controlled speech editing and zero-shot text-to-speech synthesis. It supports text replacement, insertion and deletion, emotion and prosody control, pauses, enhancement, and background-audio operations while preserving the speaker and the acoustic context outside edited regions.
Model details
The model supports English and Mandarin speech editing and zero-shot speech synthesis. It produces 48 kHz audio. The core weights use BF16; the speaker encoder and vocoder retain FP32 weights.
Supported editing operations include:
- Text replacement, insertion, and deletion
- Localized emotion control
- Pitch and speaking-rate control
- Pause insertion and reduction
- Speech enhancement and background-audio operations
- Compositions of multiple edit operations
Installation
Install the latest main branch of
dots.tts and its recommended
dependencies:
git clone --branch main --single-branch https://github.com/studio-dots-ai/dots.tts.git
cd dots.tts
python -m pip install -e . -c constraints/recommended.txt
Speech editing CLI
Source audio, a tagged instruction, and the output path are required. Source and target transcripts are optional; when omitted, both are derived from the instruction.
dots.tts.edit \
--model-name-or-path dots-studio/dots.tts.edit \
--source-audio /path/to/source.wav \
--instruction 'Hello <sub targ="small">brave</sub> world.' \
--output edited.wav
Use --source-text and --target-text to override the derived transcripts.
Source-speaker guidance defaults to auto: it is disabled when the instruction
contains at least one operation and every operation is emo, bg, or
enhance, and enabled for text, pitch, rate, pause, speaker-transfer, or mixed
edits. Pass bare --use-xvector (or
--use-xvector on) to force it on, and --use-xvector off to force it off.
Speaker guidance uses a default scale of 1.5; override it with
--speaker-scale when needed.
Python API
import soundfile as sf
from dots_tts.edit_runtime import DotsTtsEditRuntime
runtime = DotsTtsEditRuntime.from_pretrained(
"dots-studio/dots.tts.edit",
precision="bfloat16",
)
result = runtime.generate_edit(
source_audio_path="/path/to/source.wav",
instruction='Hello <sub targ="small">brave</sub> world.',
# source_text and target_text are optional overrides.
# use_xvector defaults to "auto"; pass True or False to override it.
)
sf.write(
"edited.wav",
result["audio"].float().cpu().squeeze().numpy(),
result["sample_rate"],
)
Supported structural tags include <del>, <ins>,
<sub targ="replacement">, <emo>, <pitch>, <rate>, <enhance>,
<bg>, <pause/>, and <spk_transfer/>. Malformed instructions and
instructions that derive an empty source or target transcript are rejected.
DotsTtsEditRuntime also provides generate and generate_stream for
zero-shot TTS with the same checkpoint. See the
code repository for examples.
Local Edit Playground
Build the committed frontend source with Node.js 20+ and launch the local app:
cd apps/edit_playground/frontend
npm ci
npm run build
cd ../../..
python apps/edit_playground/app.py \
--model-name-or-path dots-studio/dots.tts.edit \
--optimize
Limitations and responsible use
Generated or edited speech may contain pronunciation, timing, speaker-similarity, or instruction-following errors. Do not use the model to impersonate people, mislead listeners, violate consent, or create unlawful content. Users are responsible for complying with applicable laws and for clearly disclosing synthetic or edited audio where appropriate.
License
See the repository license metadata and the
dots.tts project for the terms
that apply to the weights and code.