SteerTTS

A small English text-to-speech model you can steer with numbers, style tags, free-text prompts and inline markup. Every clip below is raw model output, with no cherry-picking across seeds.

SteerTTS v1

Acoustic model + flow refiner + Vocos vocoder24 kHzseed 0

Default voice: v1.0.0 vs v1.0.1

“Hi, thanks for calling. I can help you reschedule your appointment today.”
voice_1361 (v1.0.0 default)
voice_1688 (v1.0.1 default)
“The weather was cold, but the view from the top was worth every single step.”
voice_1361 (v1.0.0 default)
voice_1688 (v1.0.1 default)

Steering voice_1688

“Are you coming with us tonight?”
No controls
Same text
ctrl={"f0_end": 2, "f0_iqr": 1.5}: rising, livelier
“Please keep this between the two of us.”
ctrl={"rate": -1.5, "loud": -1.5}: slow, soft
Same text
tag="whisper"
“I really thought it would work this time.”
tag="sadness", guidance=1.5
“I can't believe we actually won!”
prompt="excited, a bit faster"
Inline markup
Well, *that* was fast. [pause 400ms] <slow>Let me think about it.</slow> <whisper>Don't tell anyone.</whisper>

SteerTTS edge

11.84 MB total payloadint4 acoustic model + int4 vocoderrendered on a CPUdefault voice voice_1688

“Hi, thanks for calling. I can help you reschedule your appointment today.”
No controls
“Are you coming with us tonight?”
ctrl={"f0_end": 2, "f0_iqr": 1.5}: rising, livelier
“The weather was cold, but the view from the top was worth every step.”
ctrl={"rate": -1.5}: slower
Same text
ctrl={"rate": 1.5}: faster
“Please keep this between the two of us.”
tag="whisper"
“I really thought it would work this time.”
tag="sadness", guidance=1.5
Inline markup
Well, *that* was fast. [pause 400ms] <slow>Let me think about it.</slow> <whisper>Don't tell anyone.</whisper>

Clips are MP3 encodes of the WAV files in each Hugging Face repo.

Try it

pip install "steertts[all] @ git+https://github.com/WhissleAI/steer-tts"

steertts say "Hello there, nice to meet you." -o hello.wav

The v1 weights download from Hugging Face on first use. The repo asks you to accept the licence terms once (CC BY-NC 4.0).