ElevenLabs Instant Soundtracks scores uploaded video and places synced sound effects

What ElevenLabs launched

ElevenLabs announced Instant Soundtracks on 15 October 2026 as an update to ElevenCreative, its suite for making audio and video content. The feature takes a video you upload, scores it with original music, then analyzes the footage for moments that call for a sound effect and generates effects aligned to those timestamps. The effects are placed on the timeline automatically, so a single pass returns both a backing track and synced sound design.

ElevenLabs' own example is a whoosh timed to a logo reveal at the end of an ad. That is a small task, but it is the kind editors otherwise do by hand: scrub to the frame, find a stock effect, nudge it until it lands. The company says Instant Soundtracks is available immediately in ElevenCreative, ElevenMusic, its Creative Engine for ads, and the ElevenCreative Studio API.

How Instant Soundtracks works in the editor

In the app, the feature lives in the Music Video Editor inside ElevenCreative's music tab. According to the announcement, generated sound effects appear as separate layers from the soundtrack, so they can be reviewed and adjusted on their own. Users can add more effects, shift timing, swap out the music or refine individual moments without regenerating the whole piece, then export directly from ElevenCreative.

Keeping music and effects on separate layers is the practical detail that makes the output editable rather than a single baked-in mix. A producer who likes the score but not one effect can delete or move that effect alone. A brand team that must use licensed music can keep only the generated effects layer. The blog post does not describe how many effects the model will place per minute or how it chooses between candidate cues, so those behaviours will have to be judged in use.

Short-form timing, mood shifts and audio as a reference

ElevenLabs says the update brings better timing for short-form content, with tighter alignment around cuts, transitions and on-screen actions. It also claims more accurate mood switching across scenes. The company's example is a video that moves from a fast-paced scene, where the backing track might be EDM or R&B, to a slower, more intimate scene where the music shifts to acoustic guitar.

A second addition lets the video's existing audio act as a reference for the new score. The model can analyze that audio, including genre, crescendos, beats per minute and instrumentation, and use it to shape the generated music. ElevenLabs pitches this as producing music that matches the character of the original rather than starting from scratch. That is useful when replacing a temporary track an editor has cut to, or when a creator wants a fresh score that keeps the pacing of a draft.

All of these quality claims come from ElevenLabs. AI Lookout has not tested Instant Soundtracks, and no independent comparison against other video-to-music tools has been published.

Where you can use it today

The announcement names four entry points. The Music Video Editor in ElevenCreative offers sound effects as a separate layer and audio-as-reference controls. Creative Engine, ElevenLabs' ad-production workflow, can add music and sound effects to ads. The ElevenMusic platform also carries the feature. Developers get it through the ElevenCreative Studio API, which adds three new parameters: add_sfx, separate_tracks and use_audio_as_reference.

With separate_tracks set to true, the API returns a ZIP file containing the music and the sound effects as separate files, so teams can mix them in their own pipeline or back in Studio. ElevenLabs says each request supports up to 10 videos, combined up to 200 MB and 10 minutes of video.

What the video-to-music API accepts

ElevenLabs' Video To Music API reference documents the endpoint as POST /v1/music/video-to-music with a multipart form body. The required videos field takes one or more files that are combined in order; the documented limits match the blog post at a maximum of 10 videos, 200 MB combined and 600 seconds in total. The reference warns that combining several videos can make requests take significantly longer and suggests joining clips beforehand where possible.

Optional fields include a text description of up to 1,000 characters and up to 10 style tags such as upbeat or cinematic. The three new switches are documented plainly: add_sfx incorporates sound-effect cues from the video, separate_tracks returns a ZIP with separate music and SFX tracks instead of a single mixed file, and use_audio_as_reference analyzes the video's audio track and uses its genres, BPM, instrumentation, mood and arc as reference characteristics.

A model_id field selects the music model. ElevenLabs' public music skills reference on GitHub says the video-to-music endpoint defaults to music_v1 and that music_v2_5 is the most advanced option, so developers who want the newest model need to request it explicitly.

Formats, plan tiers and provenance signing

The API reference ties some output formats to subscription tiers. MP3 at 192 kbps requires the Creator tier or above, and 44.1 kHz PCM requires the Pro tier or above. The default output is MP3, and the content type of the response follows the chosen output_format.

There is also a sign_with_c2pa option, which signs generated songs with C2PA content credentials when the output is MP3. C2PA is an industry standard for attaching provenance metadata to media, and it is off by default. For publishers worried about disclosure, that switch is worth turning on: streaming services are starting to treat AI-generated music differently.

YouTube Music has begun labeling songs made primarily with generative AI, and Deezer says it will let listeners switch off fully AI-generated music before the end of 2026. A generated score under a short social clip is a different use from an uploaded track, but the direction of platform policy is toward more labeling, not less.

Why it matters for video and ad teams

Generative video tools are producing longer clips with fewer steps; Kling 4.0 opened to all users with 30-second video this week. Sound often remains the manual step at the end. Instant Soundtracks targets that gap by treating the finished cut as the input and returning editable music and effects, rather than asking creators to prompt audio separately and line it up afterwards.

For ad teams, the Creative Engine integration means variants of the same spot can get matched scores and effects without a sound designer on every cut. For developers, the separate-tracks ZIP fits pipelines that already mix audio programmatically, and the 10-video batch limit allows several short clips per request. Research systems have also explored joint audio-visual generation, such as Noiz AI and HKUST's Helix World, but ElevenLabs' approach is post-production: it adds sound to video that already exists.

What ElevenLabs has not said yet

The launch post does not list per-request pricing or credit costs for Instant Soundtracks, and the API reference does not state them either. It also does not say which music model powers the in-app feature, how commercial-use rights for generated scores differ by plan, or how the model handles dialogue-heavy footage where effects could clash with speech.

Things to watch include independent tests of how precisely effects land on cuts, whether use_audio_as_reference can drift too close to a reference track's melody, and whether ElevenLabs makes C2PA signing the default. Until then, the safest reading is that ElevenLabs has shipped a practical editing workflow with clear API limits, and that its timing and mood claims still need outside verification.