CapCut uses “AI voice” for three different jobs. Text to Speech creates narration from a script, Voice Changer transforms audio you already recorded, and Custom Voice builds a reusable voice profile from a consented sample. Pick the wrong tool and the menus feel confusing; pick the right one and the workflow is short.
I use this page as the decision guide: it covers what each tool takes in, where to look and what to verify before export. If you only need script-to-narration instructions, the dedicated CapCut text-to-speech guide goes deeper on long scripts, timing and platform-specific troubleshooting.
Choose the Right CapCut AI Voice Tool First
| Tool | Input | Output | Best fit |
|---|---|---|---|
| Text to Speech | Typed script or selected text layer | New narration clip | Faceless videos, explainers and fast voiceovers |
| Voice Changer | Recorded or imported speech | Altered version of that performance | Keeping your timing and delivery while changing the sound |
| Custom Voice | Your authorized voice sample | Reusable personal voice inside TTS | Consistent narration across a series |
The distinction matters for search and for the editor. A text layer leads to TTS; an audio clip leads to Voice Changer. Custom Voice may sit inside an AI Voice or My Voices area, depending on platform and rollout.
Need Scripted Narration? Use the Text-to-Speech Guide
Text to speech is a separate job from changing a recording. If you have a written script and want CapCut to generate the narration, use the CapCut text-to-speech walkthrough for mobile, desktop and web. Stay here when you already have recorded speech, want a voice effect, or need to create and test a consented Custom Voice.
Pick a Voice Without Chasing a Moving Catalog
CapCut's current marketing page advertises 200+ voices, but that number is not a promise that every account gets the same catalog. CapCut's own missing voice-options notice says server maintenance, phased rollout and outdated app or browser versions can change what appears.
I would not build a channel around a voice name from a listicle. Instead:
- Filter by the script language and, where available, accent or style.
- Preview the real hook, one name, one number and one difficult term.
- Compare two or three candidates at normal phone volume.
- Save a backup workflow in case the selected voice disappears later.
Narrator, character and effect groupings are useful browsing concepts, but names and badges move. The panel in your signed-in account is the only reliable inventory for that project.
Use Voice Changer for Recorded Speech
Voice Changer starts with audio, not text. Record yourself or import a clip, select the audio and open the voice effects area. CapCut's current Voice Changer page describes the web path as selecting the clip and choosing a voice effect under the Audio controls.
- Use the cleanest recording you can; effects magnify room echo and background noise.
- Duplicate or retain the original clip before applying a strong effect.
- Preview the whole sentence, not only the first second.
- Export a short sample and compare it with the timeline.
Voice Changer can feel more natural than TTS when rhythm and emotion matter because your performance supplies the timing. TTS is faster when you do not want to record. Neither is automatically “better”; they solve different inputs.
Create a Custom Voice Safely
CapCut promotes Custom Voice as a reusable voice profile. In accounts where it is available, open the AI Voice or My Voices area, read the notice, record the prompted sentence in a quiet room and test a short script before using it in production.
Follow the sample length shown in your account. CapCut's pages have published different sample durations, and Custom Voice availability can change by platform, region and rollout. A Pro badge or credit confirmation can also appear for some users, so check the live prompt and our CapCut AI credits guide before generating a long batch.
Consent is not a checkbox to rush through. Clone only your own voice or a speaker who understands how the result will be used. Never imitate another person or public figure without permission, and keep client approval in writing.
Fix Missing Voices, Menus, or Failed Generation
Start with the boring checks because they match CapCut's current documentation: select the text layer, verify the network, use a supported language and voice, update the app or browser, and make sure the generated audio is not muted or outside the visible timeline.
- Missing menu: select a text clip rather than an audio or video clip.
- Missing voice: refresh, reopen, update, sign out and back in, then test another network.
- Generation fails: shorten the script, remove unusual symbols and try a different available voice.
- Audio exists but is silent: check mute, volume and clip placement.
Recent community reports of Custom Voice disappearing show why I avoid promising that a menu is permanent. Those reports are directional, not proof of a universal outage. If the problem survives the documented checks, save your platform, app version and a screenshot before contacting support.
Fix Pitch or Speed That Changes After Trimming
A June 2026 community report described pitch disappearing after speed changes on web export. CapCut does not document that as a universal limitation, so test your own build rather than assuming every project is affected.
If the timeline sounds right but the export does not, apply pitch and speed to the full uncut voice clip, mute unrelated tracks, export that processed audio and reimport it before trimming. Keep the source muted until the final file passes. The reimported track has the transformation baked in instead of relying on the live effect stack.
A compound clip or a different voice effect may also work, but neither is a guaranteed fix. Render a 10-second test before processing a long narration.
Commercial Use: Check the Voice, Not Just the Tool
CapCut's current TTS page says generated audio can be used for commercial projects. For client work, I would go one step further: use the Commercial License filter when it appears and choose a voice returned by that filter. CapCut's own commercial-use walkthrough documents that control in the web TTS library.
That permission does not give you rights to somebody else's script, music, recording or identity. It also does not override the rules of YouTube, TikTok, an advertiser or a client contract. Save the selected voice screen and the applicable terms with the project when commercial rights matter.
Run an AI Voice Quality Check Before Export
My quickest useful test is one sentence containing a name, a number and an acronym. It exposes weak pronunciation before I generate a full script. Then I listen once on headphones, once through a phone speaker and once after export.
- Pronunciation: rewrite phonetic trouble spots or split the line.
- Timing: make sure pauses, cuts and auto captions still align.
- Volume: check the voice against music on a phone speaker, not only studio headphones.
- Export: listen to the delivered file; the timeline preview is not the final product.
When CapCut AI Voice Is Not Enough
CapCut is practical for Shorts, Reels, TikTok narration and explainers already being edited in the same project. For long-form work, listen for repeated cadence, flat paragraph transitions and weak control over emphasis. If the available voice cannot carry several minutes without sounding mechanical, record a human narration or compare a dedicated service such as ElevenLabs. Verify that service's current plan, consent rules and commercial terms before uploading any voice sample.
For the rest of CapCut's generators, captions and background tools, use the CapCut AI Tools hub. If a voice is unexpectedly paywalled, check the live badge against the Free, Standard and Pro pricing guide rather than relying on an old feature list.
CapCut AI Voice FAQ
How many AI voices does CapCut have?
CapCut's current TTS page advertises more than 200 voices, but the available catalog varies by language, region, platform, app version and account. Treat the signed-in voice panel as the current list.
Can I use CapCut AI voice commercially?
CapCut says its TTS output can be used in commercial projects. When available, enable the Commercial License filter and select an eligible voice. You still need rights to the script, source media and any cloned identity.
Does CapCut AI voice work offline?
No. CapCut documents text-to-speech as a cloud-processed feature, so generation needs a stable internet connection. A clip that has already been generated may remain editable in the project.
Why did a CapCut AI voice disappear?
CapCut lists maintenance, phased rollout and outdated app or browser versions as reasons voices may be missing. Refresh, update, sign out and back in, then check the voice panel again.
Can CapCut clone my voice?
CapCut offers Custom Voice in eligible accounts. Record only your own voice or a speaker who has given informed permission, follow the prompt shown in the account and test a short script before spending credits or planning a series.
What is the difference between Voice Changer and Text to Speech?
Text to Speech generates narration from typed text. Voice Changer transforms an existing recording. Use TTS when you do not want to record; use Voice Changer when your performance should supply the timing and emotion.
Which CapCut AI voice is best for TikTok?
There is no universal best voice. Preview two or three available voices with the real hook, choose the clearest pronunciation at phone volume and avoid building a channel identity around a preset that may disappear.
How do I fix pitch or speed that changes after export?
Update CapCut and export a short test. If the result still changes, render the processed voice as a separate audio file, reimport it and then trim the baked-in track. Keep the original muted until final QA passes.