AI video generators with audio, tested
Short answer: Veo 3.1 is the best AI video generator with sound, especially for speech. Kling 3.0 also makes native audio for $0.04/s less, and PixVerse V6 does it for about a third of Kling's price. Runway Gen-4.5 is silent.
Turn the sound on in the players below. Every Kling, Veo and PixVerse clip on this site was generated with audio on, and the Runway clips have no audio track.
| Model | Native audio | API price | Cost / 4s clip (as tested) | Overall score |
|---|---|---|---|---|
| Veo 3.1 Fast | Yes, native (sound effects, ambience, speech) | $0.15/s with audio, $0.10/s without | $0.60 | 7.2/10 |
| Kling 3.0 Standard | Yes, native | $0.13/s with audio, $0.09/s without | $0.52 | 7.2/10 |
| PixVerse V6 | Yes, native (generate_audio_switch in the API) | ≈$0.053/s with audio, $0.04/s without (12 / 9 credits/s at 720p) | $0.21 | 7.1/10 |
| Runway Gen-4.5 | No. Add sound separately (Runway Workflows "Add audio") | $0.12/s (silent) | $0.48 | 6.9/10 |
Prices are from the Runway Dev API (October 2026, $0.01/credit) and the PixVerse Platform API (225 credits = $1). Audio support comes from Runway's Gen-4.5 docs Google's Veo docs and PixVerse's V6 docs, checked October 16, 2026. Scores rate the picture only.
Ambient sound test: flaming wok
“Chef flipping vegetables in a flaming wok, sparks and steam, kitchen ambient sound.” The prompt asks for kitchen sound.
adherence 9motion 8quality 8artifacts 6
Real toss with vegetables airborne and flames; some morphing between beats.
adherence 7motion 7quality 9artifacts 8
Vivid flames and sparks, close-up; stirring rather than a true flip.
adherence 9motion 8quality 8artifacts 6
Big flame burst and vegetables tossed into the air; some morphing between beats.
adherence 6motion 6quality 8artifacts 8
Wide kitchen shot with a flame burst; no visible vegetable flipping.
Dialogue test: park bench
“Two people having a conversation on a park bench; she says "I told you it would work.".”
adherence 8motion 6quality 8artifacts 8
Close two-shot, she clearly speaks; natural expressions.
adherence 7motion 6quality 8artifacts 7
Close two-shot with natural expressions, but her line is hard to see being spoken.
adherence 6motion 5quality 7artifacts 7
She speaks briefly, then both freeze; stiff.
adherence 6motion 6quality 7artifacts 7
Wide shot on bench, mostly smiling; little visible dialogue.
More detail: AI video lip-sync test.
Which to pick
- Speech, talking heads, UGC-style ads: Veo 3.1. Google Veo
- Action and food with ambient sound: Kling 3.0. Kling
- Lots of clips with sound on a tight budget: PixVerse V6 at $0.21 per 4-second clip (review). PixVerse
- You'll add music in editing anyway: Runway Gen-4.5, the cheapest model on the Runway Dev API. Runway
Higgsfield also offers audio-capable models. We haven't tested it yet. It joins our November run (Higgsfield *; * affiliate link, disclosure).
Related: Sora alternatives · Cheapest AI video generator · Cost per second · Best for TikTok & Shorts
FAQ
Which AI video generator makes video with sound?
Of the models we tested, Veo 3.1, Kling 3.0 and PixVerse V6 generate audio together with the video. Runway Gen-4.5 is silent; Runway's own docs say audio has to be added separately. Kling 4.0 (launched October 16, 2026, not tested yet) also lists audio.
How much extra does audio cost?
Through the Runway Dev API in October 2026, audio raised Veo 3.1 Fast from 10 to 15 credits per second and Kling 3.0 from 9 to 13 ($0.01 per credit). For a 4-second clip that's +$0.20 on Veo and +$0.16 on Kling. On the PixVerse Platform API, audio takes V6 from 9 to 12 credits per second at 720p (225 credits = $1), about +$0.05 per 4-second clip.
Which has the best dialogue and lip sync?
Veo 3.1 Fast. In our park-bench prompt where a character says a specific line, Veo scored 7.5 and clearly showed her speaking. PixVerse V6 scored 7.0 (a close two-shot, but her line is hard to see being spoken), Kling 6.3 (she speaks briefly, then both freeze) and Runway 6.5 (little visible dialogue, no audio).
Did you score the audio itself?
Not separately. Our October scores rate the picture: prompt adherence, motion, quality and artifacts. We confirmed which clips contain an audio track, but there is no audio score yet. We plan to add one.