A silent AI clip usually means one of four things: the model has no sound setting on 8frame, the sound was switched off (and on Kling O3 it starts off), a reference audio track made the model drop its own, or the model was never meant to make sound because it takes your audio instead. The fix depends on which one it is, and three of the four are settings rather than failures.
TL;DR
- Some models have no sound setting on 8frame: Hailuo H3, Happy Horse, base Wan 3.0, Wan 2.5, Wan 2.2 and Runway Gen-4 Turbo
- Some let you switch it off: Veo 3.1 and Kling v3 and 2.6 Pro have it on by default; Kling O3 starts with it off
- Seedance drops its own sound when you attach a reference audio track
- Lip sync tools take your audio, they do not generate any
- No sound on the model you need? Add it after: a voiceover, a music bed or sound effects on the same canvas
1. The model has no sound setting
These video models have no generated-audio setting on 8frame: Hailuo H3, Happy Horse, base Wan 3.0, Wan 2.5, Wan 2.2 and Runway Gen-4 Turbo. For base Wan 3.0, Alibaba's own API generates sound by default, but we have not verified what our integration returns, so do not rely on it.
Fix: choose a model that generates sound, or add sound after (section 5). The models that do, and what sound costs on each, are in AI video models with audio, by model.
2. Sound was switched off
On Veo 3.1 (all tiers), Kling v3 and Kling 2.6 Pro, sound is on by default and switching it off lowers the price, so it is easy to have turned it off for drafts and forgotten. Kling O3 works the other way: sound is off until you switch it on.
Fix: check the audio setting on the node and regenerate. Expect the price to rise: on Veo 3.1 Lite an 8-second clip is 33 credits silent and 54 with sound; on Kling O3 a 5-second clip is 57 and 76.
3. A reference audio track turned it off
Seedance 2.5 and 2.0 generate sound by default, but switch their own sound off when you attach a reference audio track, so your soundtrack is not layered with a generated one. If you attached a track to guide the clip, the output carries no generated sound.
Fix: remove the reference track if you want Seedance's own sound, or keep it and add your audio in the edit.
4. The tool takes your audio instead
Lip sync tools (Pruna Avatar, Hailuo H3's lip sync mode, OmniHuman) animate a face to audio you provide; they do not generate sound of their own. Happy Horse's video edit can keep the original clip's sound, but it does not create new sound.
Fix: generate the voice first, then feed it in. On 8frame, ElevenLabs text to speech is 13 credits per started 1,000 characters.
5. Adding sound when the model has none
For a clean mix this is often the better route anyway: generated sound arrives as one baked track that you cannot rebalance. Build the sound separately and lay it under the picture:
| Sound | On 8frame | Credits |
|---|---|---|
| Voiceover | ElevenLabs TTS | 13 per started 1,000 characters |
| Sound effect | ElevenLabs SFX | 1 to 6, by length |
| Music bed, 30 s | Lyria 2 | 13 |
| Music with lyrics, 60 s | MiniMax Music 3 | 17 |
If the model should have made sound and did not: download the file and play it in a desktop player before assuming it is silent. If it is silent there, check the node's audio setting and the model's row above before regenerating.
Pruna and Creatify Boreal
Pruna p-video, p-video 2 Pro and Creatify Boreal always add sound on 8frame and have no silent setting. The one we cannot vouch for is Pruna p-video Draft: the provider saves sound by default, but we have not confirmed that draft output keeps it.
FAQ
Why is my Hailuo H3 video silent? Hailuo H3 has no generated-audio setting on 8frame. Add sound after, or pick a model that generates it.
Why is my Kling O3 video silent? Kling O3 starts with sound switched off. Switch it on before generating; a 5-second clip goes from 57 to 76 credits.
Why did Seedance stop generating sound? Seedance switches its own sound off when you attach a reference audio track.
Which AI video models always include sound? On 8frame: Grok Imagine, Gemini Omni Flash, Creatify Boreal, Pruna p-video and p-video 2 Pro. None of them can be made silent.
Sources
- Audio settings, defaults and prices per model: the settings 8frame sends to each provider and the 8frame tool registry, read 2026-09-30
- Pruna sound: the provider's input schema and example outputs for prunaai/p-video and prunaai/p-video-2-pro, checked 2026-09-30
- Wan 3.0 sound in Alibaba's API: Wan 3.0 video generation guide, checked 2026-09-30
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.