Audio
When to keep a model's native audio, and how to add voice, music, sound effects and lip sync without license surprises.
Keep the model's native audio for short, throwaway clips. For anything longer, cut from several shots, or made for a brand, lock the picture first and then add voice, music and sound effects as separate layers. Audio is often the first thing that makes a video feel cheap, and free plans of most audio tools do not allow commercial use.
Native audio or a separate pass
| Situation | Use |
|---|---|
| Single clip, meme, test | Native audio from the video model |
| Several shots cut together | Separate pass: shot to shot native audio rarely matches |
| Ads, client work, anything you will revise | Separate pass, so you can change the music or a line without regenerating video |
| Dialogue on camera | Native audio for a draft, then a dedicated lip sync step if it drifts |
How the models compare (vendor blogs and user reports, no blind tests):
- Veo 3.1 has the most natural native audio, according to several vendor comparisons.
- Kling 3.0 native audio is described as functional. One user said Kling's generated music sounds like an older version of Suno.
- Seedance 2.5 invents sound or ignores the audio reference in roughly 15 to 20% of runs, according to one creator.
- Wan 3.0 has strong visuals but weak audio, according to users.
Whatever you choose, write the audio into the prompt. Every native audio model fills an empty audio slot, usually with music. A common habit is to ask for sound effects only and add music in the edit.
Audio: rain on window, low fridge hum, single cup set down at 3s. No music, no voices.A workflow that works for most projects:
Prototype with native audio
It tells you whether the timing works.
Lock the picture
Finish the edit before you spend on audio.
Lip sync, if needed
Do it as its own step on the final cut of each talking shot.
Layer voice, music and sound effects
Use dedicated tools for each, then mix.
Voice
ElevenLabs is still the quality reference for AI voice in 2026.
| ElevenLabs plan | Price | Credits per month | Commercial use |
|---|---|---|---|
| Free | 0 USD | 10,000 | No |
| Starter | 6 USD/month | 30,000 | Yes |
| Creator | 22 USD/month | 121,000, plus professional voice cloning | Yes |
| Pro | 99 USD/month | 600,000 | Yes |
Prices checked on 15 Oct 2026 on the official pricing page. Annual billing gives 2 months free.
Our pick: Starter at 6 USD is enough for voiceover on short videos and gives you a commercial license. Move to Creator only if you need a professional clone of a specific voice.
Cheaper options, all from third party pages with conflicting prices:
- Fish Audio: the most cited cheaper option, about 5 to 15 USD per 1 million characters. Sources disagree on whether its free tier allows commercial use.
- Hume Octave 2: emotional reads, about 7.60 USD per 1 million characters.
- Kokoro: free, Apache 2.0, runs on your own computer.
YouTube does not require an AI label when you clone your own voice for a voiceover.
Music
| Tool | Price | Commercial use | Notes |
|---|---|---|---|
| Suno Free | 0 USD, 50 credits a day | No | v6 mini model, no downloads |
| Suno Pro | 8 USD/month (billed yearly) | Yes | 2,500 credits, 20 downloads a month |
| Suno Premier | 24 USD/month (billed yearly) | Yes | 10,000 credits, 60 downloads a month, Suno Studio, stems |
| ElevenLabs Eleven Music | Included in ElevenLabs plans, about 900 credits per minute of music | Yes from Starter; self serve terms exclude film, TV, radio and studio games | Trained under licensing deals with Merlin and Kobalt |
| Udio | Downloads disabled since the 2025 UMG settlement | Unsettled | A licensed platform was announced; launch status unclear |
| Pixabay, Freesound | Free libraries | Check each track's license | Good for the smallest budgets |
Prices checked on 15 Oct 2026 (Suno and ElevenLabs official pages).
The Suno free plan trap
Songs you make on Suno's free plan never become commercially usable, even if you upgrade later. Songs made while you are subscribed keep their rights after you cancel. If a song might end up in a monetized video, make it on a paid plan from the start.
Our pick: Eleven Music has the cleanest rights for ads and client work. Suno Pro is better when you want stems or more control over the arrangement. Either one can still trigger a YouTube Content ID claim (single source). On YouTube, a realistic video where AI music is the main focus needs the AI disclosure.
Sound effects
Sound effects make AI footage feel real faster than any other layer. You have two kinds of tools: ones that watch your video and generate matching sound, and ones that make a sound from a text prompt.
| Tool | How it works | Price |
|---|---|---|
| Mirelo SFX 2.0 | Watches the video and generates footsteps, impacts, foley and ambience. Can remove unwanted baked in music while keeping dialogue. Videos up to 10 minutes | Free: 5,000 credits (about 8 minutes), watermark, projects deleted after 1 week. Creator 20 EUR (about 40 minutes). Studio 99 EUR (about 200 minutes) |
| ElevenLabs SFX v2 | Text to sound effect, up to 30 s, 48 kHz, can loop | About 200 credits per generation, inside your ElevenLabs plan |
| Freesound, Pixabay | Free libraries | Free |
Prices checked on 15 Oct 2026.
Notes before you pay:
- Mirelo SFX 2.0 was released on 14 Oct 2026. It has plugins for Premiere and Resolve, an MCP server for Claude and an API. Its pricing page does not state commercial license terms, so read the terms of service. The launch was pushed by a coordinated wave of posts on X, and the blind test numbers in those posts (57% vs 21% against Veo 3.1 Fast) came from the vendor.
- ElevenLabs says free plan sound effects need attribution and paid plans need none. Third party guides conflict, so use a paid plan for commercial work.
- ElevenLabs video to sound is labeled an experiment.
Lip sync
Short lines lip sync best. Long speeches drift.
| Tool | Best for | Price |
|---|---|---|
| Hedra (Character-3) | Stylized or seated characters talking to camera; cheaper and faster | Basic 20 USD/month (2,000 credits), Pro 50 USD (7,200), Ultra 100 USD (18,000), Teams 75 USD. About 6 credits per second, so Basic covers roughly 5 minutes. Commercial use on all paid plans |
| Kling lip sync | Characters made in Kling, multilingual dubbing | Extra credits, on Pro and higher plans |
| Sync.so | Real camera footage | Check the pricing page |
| HeyGen | Presenter avatars | API access from 100 USD/month |
Prices checked on 15 Oct 2026. Hedra raised prices in September 2026 (Basic was 15 USD).
How to choose (vendor comparisons, so treat as a direction only):
- Character made in Kling, or you need several languages: Kling.
- Stylized character talking to camera: Hedra.
- Real person on real footage: Sync.so.
- Presenter style talking head: HeyGen, which a secondhand summary of Reddit calls the standard.
VEED, a competitor, claims Kling shows a "floating mouth" artifact. Tips that help with any tool:
- Keep lines short and split long speech across cuts. One reviewer saw drift past about 45 seconds.
- Use a frontal, high resolution face as the source image.
- Supply your own voice track (for example from ElevenLabs) instead of a generated one when timing matters.
- With Seedance, one creator puts the voice track on a black video reference cut to the clip length, plus an image reference. The same creator finds MiniMax H3 Max easier for audio guided lip sync.
Mixing basics
- Put dialogue, foley, ambience, sound effects and music on separate tracks or buses, so you can adjust each group.
- In Resolve, the Audio Assistant gives a first pass mix (leveling, voice isolation, de-essing). Treat it as a proposal and listen through.
- Keep music under the voice. When someone speaks, the words win.
- For TikTok, Reels and Shorts, third party guides suggest about minus 14 LUFS for the final loudness. See Finishing for export settings.
- Listen once on phone speakers and once on headphones before you export.
License checklist
- Free tiers of ElevenLabs (voice and music), Suno and Kling carry no commercial rights.
- Rights attach to output made while you are subscribed.
- Keep a note of which plan you were on when you made each track.
Sources
- ElevenLabs pricing, Eleven Music terms, sound effects commercial use
- Suno pricing and Daily AI Fixs, Suno free songs stay non commercial
- Dubspot, AI music licensing
- License Org, Suno vs ElevenLabs licensing
- Mirelo and Mirelo pricing
- Hedra pricing and Costbench, Hedra price change
- VEED, lip sync API comparison
- BudgetPixel, native audio vs adding sound later
- OriSilver on X, audio driven lip sync
- YouTube, altered or synthetic content