ElevenLabs released Studio 4.0 on September 21, 2026, the new version of the editor inside ElevenCreative, its platform for voice, music, image, and video generation. The change in one sentence, from the company’s announcement: generate video, image, voice, music, and sound effects inside a project, then edit, caption, and export in the same place.
This article is for people who make videos, podcasts, or audiobooks and want to know what changed, which models do the generating, which plan they need, and what a finished video costs. The first three questions have answers below. The fourth does not, because ElevenLabs prices generation in credits that vary by model and shows the estimate only inside the interface. We have not tested the release; details come from ElevenLabs’ blog post, product page, documentation, and pricing page. The documentation still describes Studio 3.0 in places.
What is new in Studio 4.0
- Generation inside the project. Video, image, voice, music, and sound effects are created from within a Studio project and land on the timeline, instead of being generated in a separate playground and imported. Every model available in ElevenCreative can be used from the editor.
- Refine a shot in place. A single generated shot can be regenerated or adjusted without rebuilding the sequence or re-uploading.
- Captions as clips. Each caption sits on the timeline and is trimmed and moved like any other clip.
- A rebuilt timeline for multi-scene, multi-track projects such as ad campaigns, tutorials, and social cuts: frame-level zoom, clip snapping, and grouped moves that keep tracks in sync. ElevenLabs also says the playback engine was rebuilt.
- Review and commenting. Comments attach to a specific clip or asset. Documentation describes shared links, timestamped comments, and threaded replies; simultaneous editing by several people is not described.
- Studio Agent as co-editor, covered below.
Studio 3.0 already had the timeline, uploaded video, auto captions, voices in 32 languages, and shared links with comments, so 4.0 is about generation moving inside the project and the timeline handling more complex edits. Our ElevenLabs voiceover guide covers the audio workflow, which has not changed.
Studio Agent: what it is and where it does not work
Studio Agent, first released in May 2026, takes a text description and edits the timeline: it drafts a first cut by placing clips, generates voiceovers, and syncs sound effects to footage, analyzing uploaded clips frame by frame so an instruction such as placing a sound when a logo appears needs no manual timestamp. You can take over manually at any point and hand the timeline back; the agent resumes from where you left off. When it generates visuals, it picks from five preset image models and five preset video models.
Three limits from the documentation matter before you plan a workflow around it. It works only in video projects, not audio-only ones. It is available on web only, with no API access. And it is billed twice: the conversation itself in tokens across input, output, thinking, cache read, and cache write, and any speech, image, music, sound, or video it generates in standard credits on top.
Which models are behind the generate button
ElevenLabs trains its own voice, music, and sound models. Video and image generation route to third-party models, and the choice sets both quality and credit cost.
| Type | Models listed in ElevenLabs docs | Notes |
|---|---|---|
| Video | Google Veo 3.1 (Fast, Lite), Gemini Omni Flash 1.1, OpenAI Sora 2 and Sora 2 Pro, Runway Gen-4.5, Gen-4 Turbo, Aleph, Act-Two, Black Forest Labs FLUX 3, LTX 2, Wan 3.0 and 2.6, Kling 3.0 and O3, ByteDance Seedance 2.0 and 2.5, MiniMax H3 | Kling, Seedance, and MiniMax are marked unavailable in the United States. Output up to 4K depending on model |
| Image | OpenAI GPT Image 2.5 and 2, Google Nano Banana Pro and 2, FLUX.2 Pro, FLUX.1 Kontext, Recraft V4.1, Krea 2, Runway Gen-4 Image, Seedream 5 | Seedream unavailable in the United States |
| Lip sync and utility | HeyGen Avatar 4, Sync 3, Veed Lipsync 2.0, Creatify Aurora, Topaz Upscale, background removal | |
| Audio | ElevenLabs voices (10,000 plus in the library, 32 languages), Eleven Music, sound effects, Voice Isolator, Speech Correction | First-party |
Plans: who gets what
| Plan | Monthly list price | Credits per month | Studio projects | Commercial license | Video export | Audio export |
|---|---|---|---|---|---|---|
| Free | $0 | 10,000 | 3 | No | Watermarked; no video generation | 128 kbps MP3, or WAV made from a 128 kbps source |
| Starter | $6 | 30,000 | 20 | Yes | Watermark-free | Same as Free |
| Creator | $22 (first month discounted to $11) | 121,000 | 1,000 | Yes | Watermark-free | 128 and 192 kbps MP3, or WAV from a 128 kbps source |
| Pro | $99 | 600,000 | 3,000 | Yes | Watermark-free | 16-bit 44.1 kHz WAV, 192 kbps MP3 |
| Scale | $299 | 1,800,000 | 20,000 | Yes | Watermark-free | Same as Pro; 3 workspace seats |
| Business | $990 | 6,000,000 | 20,000 | Yes | Watermark-free | Same as Pro; 10 workspace seats |
Sources: ElevenLabs’ pricing page and Studio documentation. The two disagree on Free projects (the pricing page says 3, the documentation says 5); the table uses the pricing page. Free accounts get three image generations a day and cannot generate video. Starter is the threshold for commercial use and clean video export; Pro is the threshold for full-quality WAV.
What a video costs: not answerable from public information
Per-generation pricing is difficult to evaluate because ElevenLabs shows the credit estimate inside the interface, one generation at a time, rather than publishing a comparable rate table. The cost depends on the model, resolution, and duration, and Studio Agent adds token billing on top. We could not calculate how many clips each plan buys, or how a 10-second clip on Veo 3.1 compares with the same clip on Sora 2 Pro, without testing the product. Until ElevenLabs publishes rates or we run that test, treat the credit allowances above as budgets of unknown size. Our comparison of AI video editing tools covers editors whose generation pricing is published.
Documented, not documented
| Question | Status |
|---|---|
| Is Studio Agent available in audio-only projects, or via the API? | No on both, per documentation. Web only, video projects only |
| What does a generated video clip cost in credits? | Not published as a rate. An estimate appears in the interface before you confirm |
| Do the 4.0 timeline features apply to existing 3.0 projects? | Not addressed in the announcement or blog post |
| Are caption clips exported as a separate subtitle file or burned in? | Documentation describes MP4 with burned-in captions and caption export in 32 languages; whether both apply to 4.0 caption clips is not stated |
| Can several people edit one project at the same time? | Not described. Documentation covers shared links, timestamped comments, and threaded replies |
| Which regions get which video models? | Documentation marks ByteDance and MiniMax models unavailable in the United States; no full regional list |
What to do
- Already using Studio 3.0: open an existing video project and look for the generate options and caption clips. Nothing indicates a migration step.
- New to it: the Free plan is enough to test the editor, captions, and voiceovers. Video generation and clean exports start at Starter.
- Producing for clients: Starter covers commercial use and watermark-free video. Before committing to a plan, generate one representative clip on each model you might use and record the credit estimate; that is the only way to price a project.