#the voices I generate have some pauses, is there a way to remove these small pauses? (I want fluid

1 messages · Page 1 of 1 (latest)

opaque lotusBOT
#
AI-generated Answer for _.gojo

Yes—there are a few documented ways to minimize or eliminate unintended micro‑pauses:

  • Remove pause‑inducing markup/punctuation in your input: don’t include SSML <break> tags, and avoid dashes/em‑dashes or ellipses, which can cause the model to insert short pauses. (help.elevenlabs.io)
  • Increase the Speed setting slightly (>1.0) to tighten timing. Speed is available in Text to Speech on the website/Studio/Conversational AI and via the API (range 0.7–1.2). Programmatically, set the voice’s speed parameter using the Edit Voice Settings endpoint. (help.elevenlabs.io)
  • If you’re using a cloned/custom voice, check your training audio. Voices trained with fillers like “uh/ah” or with choppy phrasing can reproduce those mannerisms; use cleaner, longer continuous samples to improve pacing. Also note that pacing is influenced by the audio used to create the voice. (help.elevenlabs.io)
  • Write prompts in a natural, continuous narrative style to encourage smoother delivery. (elevenlabs.io)

These steps should produce more fluid speech without the “weird” small pauses.

Sources:
How can I add pauses? – ElevenLabs
Can I change the pace of the voice? – ElevenLabs
How can I add pauses? – ElevenLabs
Controls | ElevenLabs Documentation

vale pilot
#

I'm using Liam's voice, american, social media, and the generated voices have pauses even if I remove all pause inducing markups/punctuations in my input...

opaque lotusBOT
#
AI-generated Answer for _.gojo

Thanks for the details. Even without punctuation, voices will still apply natural prosody; some voices may add brief pauses or fillers based on how they were trained. ElevenLabs notes that different voices can handle pauses differently and that some may insert “uh/ah”-like mannerisms during pauses. (elevenlabs.io)

What you can do to minimize those pauses and get a more fluid read:

  • Increase Stability for the Liam voice. Higher Stability reduces randomness between generations and yields more consistent delivery (trade‑off: can sound more monotone). (elevenlabs.io)
  • Increase the Speed setting (e.g., toward 1.1–1.2). Speed control is available across Text to Speech, Studio, and the API; valid range is 0.7–1.2. (help.elevenlabs.io)
  • Try a faster model. Flash v2.5 and Turbo v2.5 are designed for low‑latency, steady pacing; switching models can change delivery characteristics. (elevenlabs.io)

API example (adjust to taste):

  • Endpoint: POST /v1/voices/:voice_id/settings/edit
  • Body: { "stability": 0.9, "speed": 1.15 }
    These fields are officially supported voice settings. (elevenlabs.io)

If you’re already doing all of the above and still hear unwanted gaps, please share a short example line you’re using and which model you selected; I can test with those exact settings.

Sources:
Controls | ElevenLabs Documentation
Get voice settings | ElevenLabs Documentation
Can I change the pace of the voice? – ElevenLabs
Models | ElevenLabs Documentation
Edit voice settings | ElevenLabs Documentation

vale pilot
#

so its either changing the stability/speed or model (which one is better for no pauses?) otherwise changing the voice in itself because some voices were trained with the pauses, did I understand you correctly?

opaque lotusBOT
#
AI-generated Answer for _.gojo

Mostly correct — with two nuances from the docs:

  • Speed: This is the only global setting that directly compresses timing. Increase the Speed setting (>1.0) to tighten delivery across all models; range is 0.7–1.2, with very high values potentially degrading quality. (help.elevenlabs.io)

  • Model choice: No help‑center page claims a specific model “has no pauses.” If you need explicit control over pacing/pauses, Eleven v3 adds Audio Tags that let you steer rhythm (e.g., [RUSHED]) and even pause behavior, so it’s the model with the most control for fluid delivery. Note v3 is an alpha, higher‑latency model; use it when you can afford offline/non‑realtime generation. (help.elevenlabs.io)

  • Voice choice: Yes — the voice matters. Eleven’s docs note some voices trained with natural filler sounds (“uh/ah”) may insert those during pauses; if a voice tends to hesitate, switching voices can help. (help.elevenlabs.io)

  • Stability: You can try raising Stability for more consistent deliveries, but it’s not documented as a pause‑removal control. Lower Stability increases variation between generations. (help.elevenlabs.io)

Practical order of operations (per docs): 1) increase Speed slightly, 2) if possible use Eleven v3 with a pacing tag like [RUSHED], 3) try a different voice if the current one exhibits filler mannerisms. (help.elevenlabs.io)

If you want, tell me exactly where you’re generating (API, Studio, or Agents), and I’ll point to the precise parameter or UI control. (help.elevenlabs.io)

Sources:
Can I change the pace of the voice? – ElevenLabs
What is Eleven v3 (Alpha)? – ElevenLabs
How can I add pauses? – ElevenLabs
Why does my voice sound different in the API compared to the website? – ElevenLabs
Can I change the pace of the voice? – ElevenLabs
Can I change the pace of the voice? – ElevenLabs

vale pilot
#

then which fluid voice do you suggest for story naration without pauses