Likely causes and fixes (based on ElevenLabs docs)
If you’re chunking text via the TTS API, you need Request Stitching to keep prosody continuous across chunks. Without it, joins can sound abrupt. Note: Request Stitching is not supported on the eleven_v3 model. Use a supported model (e.g., multilingual_v2, turbo_v2.x, flash_v2.x) and follow the stitching guide. (elevenlabs.io)
If you’re adding pauses, overusing SSML break tags can introduce artifacts or strange pacing that can feel like bad “stitches.” Use <break time="..."/> sparingly (up to 3s). For eleven_v3 specifically, use [pause], [short pause], or [long pause] instead of SSML breaks. (help.elevenlabs.io)
If you’re working in Studio, continuous audio is rendered from Export (chapter or whole project). Export also triggers Auto‑Regenerate, which can automatically fix mispronunciations/artefacts and improve the final render. Make sure you’re exporting (not just previewing) to produce a single, properly rendered file. (help.elevenlabs.io)
If none of the above matches your setup, tell me:
Where you’re generating (API vs Studio vs Dubbing)
The model you’re using
Whether you split text into chunks
A project or request ID
Or contact Support with those details so we can inspect your project/logs. (help.elevenlabs.io)
pliant garnet
weary osprey