Product

ElevenLabs' November Updates for Video Teams: Sound Effects v3, Clone Fine-Tuning on v4 and a Video-to-Music API

9 November 2026 · RenderBob team

ElevenLabs shipped sound-effects v3, voice-clone fine-tuning on v4, and a Video-to-Music API between 19 October and 2 November. Notes come from a third-party feed.

ElevenLabs shipped a batch of audio updates between 19 October and 2 November. Three matter to video teams: a new sound-effects model, voice-clone fine-tuning on its v4 speech model, and an API that scores a video. The details come from a third-party release-notes feed.

What the feed lists

Releasebot's ElevenLabs feed, which we read, gives these dated entries. We have not checked them against ElevenLabs' own changelog.

• 2 November: SFX v3. The new model, eleven_text_to_sound_v3, is in every workspace. It follows multi-part prompts more closely, handles music and vocals better and is less likely to clip on high-frequency sounds. v2 is deprecated and will be removed on 1 March 2027.
• 2 November: voice clones. Professional Voice Clone fine-tuning for Eleven v4 and v4 Turbo is available to all workspaces. Clones fine-tuned on v4 use their saved clone-strength setting by default.
• 19 October: Video to Music API. Upload up to 10 videos (200 MB combined) and get a generated soundtrack, with optional style tags, sound effects and separate stems.
• 19 October: other changes. Zero Retention Mode for music and sound effects, WAV output for sound effects, and longer Speech Engine conversations of up to 40 minutes.
• Voice agents. Eleven Realtime reached every workspace on 2 November, completing a rollout that began on 19 October. We covered the realtime models on 22 October.

What this means for editors

The feed describes features, not quality. Sound effects and scores are only useful if they hold up in a mix, so the practical tests are about control and records.

• Check the deprecation date if a pipeline calls the v2 sound-effects model.
• Test stems. Separate stems matter more than a finished track when you need to mix.
• Check the retention setting for unreleased material.
• Check licence and usage terms for generated music and sound effects.
• Record the model and version for each cue.
• Review any voice clone for consent and keep the permission on file.

What is not shown

The feed gives no samples, pricing or limits beyond those above, and no independent testing. It does not say which plans can use each feature.

What to watch

ElevenLabs' own changelog, pricing for the new API, independent comparisons of scored output, and further changes to model deprecations.

RenderBob 3 and this story

RenderBob 3 keeps sound decisions with the picture they belong to: generated clips sit on their own timeline lane, the model is chosen per clip, and references and prompt logs stay with the studio.