ElevenLabs adds distance and reverb to generated speech

A character stands across a large room, but the voice sounds as close as a podcast microphone. ElevenLabs’ October 12, 2026 API update gives creators controls for placing generated speech within a scene: distance, reverb and stereo panning, alongside filter presets and background noise. Original release notes.

Make the voice fit the shot

Text to Speech and Text to Dialogue accept an optional audio_effects object. The release points developers to GET /v1/audio-effects/catalog for valid preset, environment and background-noise IDs. Speech reference, dialogue reference.

An illustrative first comparison is a single line heard from across the room. Keep a clean version and an effected version, then compare both with the picture and existing soundtrack. Listen for whether the voice belongs in the scene and whether every important word remains understandable. A convincing room sound that hides the line would fail that check. We have not generated or listened to this example.

For an explanatory voiceover, closeness may be exactly what you want. Keep the clean version available instead of applying the same effect to every line. Ordinary effects in your editor remain a useful option when you already have the recording.

The output format can stop the request

The endpoint references say reverb with a positive send level and non-zero panning produce stereo. Those effects are rejected for ulaw_*, alaw_* and pcm_* outputs. Unknown effect IDs also produce an error. Check the catalog and your downstream encoding before sending a full script. Documented restrictions.

ElevenLabs says effects do not change a request’s cost. Normal generation usage still applies; its pricing explanation describes model-dependent credits and per-generation charging. A second rendered attempt is still something to account for in your workflow.

This is a documented API change. Our ElevenLabs voice guide covers the wider narration workflow, while Eleven v4 and Turbo coverage separates content narration from live voice work. Start with the one cue whose relationship to the picture needs improvement.