Ovi alternative
Ovi alternative: talking clips with sound, straight from the browser
Pixel Dojo runs Veo 3.1, Seedance 2.5, Kling 3.0 and WAN 3.0, four engines that write picture, voice and sound together, with the credit price shown before you click and no graphics card to buy.
Ovi made a splash as an open model that generates video and audio in one pass. We wanted to know if you need to run it yourself to get that, so we wrote one dockside line with gulls and a foghorn and gave it to four engines we carry.
36 credits ($2.25) per 8 s clip at 1080p with audio. Included in every plan, from $10/month.
Already a member? Sign in
By Brian Lovett · Updated
More clips with sound, made on Pixel Dojo
Every clip below came from the prompt on its tile. Tap one to start from the same prompt.
Two people in one shot on Veo 3.1 Lite, with wind and the clink of a cup, for 9 credits.
Two hikers in their forties sit on the bench of a stone mountain hut at sunset, a man pours tea from a steel thermos and hands a cup to a woman, she grins and says, "Worth every step." Wind across the ridge, the clink of a cup.
Text plus image, the other way Ovi works: we made the still first, then Kling 3.0 animated it with the music in the sound track.
The accordion player squeezes the bellows and plays a slow lilting tune, he sways gently and smiles at a passerby, pigeons hop around his feet and the fountain splashes behind him, the camera holds steady. Sound: an original accordion melody, fountain water and pigeons cooing.
Pure sound design on Kling 3.0: fire crackle, rising steam and rain, no dialogue at all.
Close up of a dented copper kettle on a cast iron wood stove in a cabin as rain streaks the window behind it, steam builds and the kettle starts to whistle, the fire crackles in the stove, warm amber light, shallow depth of field.
The start frame for the accordion clip, a street photo on Nano Banana 2.1 for 1.5 credits.
Photograph of a street accordion player in his seventies with a flat cap and a white mustache, sitting on a stool in a sunny cobbled square beside a fountain, pigeons at his feet, warm afternoon light, 50mm lens, natural color.
The usual advice: self host Ovi because the weights are free
Ovi is open. Its GitHub page lists an Apache-2.0 license, authors from Character AI and Yale, and a model that makes video and audio together from text or from text plus an image. Free weights sound like free clips.
The same page sets the bill. Ovi needs at least 32GB of GPU memory, or about 24GB with an fp8 version that loses a little quality. Clips are 5 seconds, or 10 seconds since Ovi 1.1 in November 2025, at 720 by 720 or 960 by 960 and 24 frames a second.
So the real cost is a large graphics card, setup time and one model. We think most people who want a talking clip are better served by four engines in a browser tab, which is what we tested next.
One dockside line, four engines that make their own sound
We gave Veo 3.1 Fast, Seedance 2.5, Kling 3.0 and WAN 3.0 the same prompt: a fisherwoman coils a rope on a foggy pier, looks up and says one line, with gulls, waves and a foghorn. Each tile shows its settings and credits.
Medium shot of a fisherwoman in her fifties in a yellow oilskin coat on a foggy wooden pier at dawn. She coils a wet rope, looks up at the camera and says, "The fog lifts by nine. It always does." Gulls cry, waves slap the pilings, a distant foghorn.
Medium shot of a fisherwoman in her fifties in a yellow oilskin coat on a foggy wooden pier at dawn. She coils a wet rope, looks up at the camera and says, "The fog lifts by nine. It always does." Gulls cry, waves slap the pilings, a distant foghorn.
Medium shot of a fisherwoman in her fifties in a yellow oilskin coat on a foggy wooden pier at dawn. She coils a wet rope, looks up at the camera and says, "The fog lifts by nine. It always does." Gulls cry, waves slap the pilings, a distant foghorn.
Medium shot of a fisherwoman in her fifties in a yellow oilskin coat on a foggy wooden pier at dawn. She coils a wet rope, looks up at the camera and says, "The fog lifts by nine. It always does." Gulls cry, waves slap the pilings, a distant foghorn.
- Veo 3.1 Fast ran 6 seconds and gave the cleanest performance: rope first, then she looks up and talks straight to camera.
- WAN 3.0 turned the sound cues into picture, with gulls overhead and waves at the pilings, for 10 credits.
- Seedance 2.5 had the thickest fog and the moodiest frame of the four.
- Kling 3.0 framed her closest at the bollard, the version we would pick for a vertical crop.
Ovi vs Pixel Dojo: what each takes to run
Everything in the Ovi column comes from the Ovi GitHub README, read on October 15, 2026.
| At a glance | Ovi | Pixel Dojo |
|---|---|---|
| What you need | A GPU with at least 32GB of memory, or about 24GB with the fp8 version | A browser; everything runs in the cloud |
| Engines | One model; its video branch starts from WAN 2.2, per its README | Veo 3.1, Seedance 2.5, Kling 3.0, WAN 3.0 and more, all with sound |
| Clip size | 5 or 10 seconds at 720 by 720 or 960 by 960, 24 frames a second | Our test clips ran 5 and 6 seconds at 720p and 1080p in 16:9 |
| Writing speech | Speech wrapped in special tags, sound described after "Audio:" | Write the line in quotes inside a plain prompt |
| Cost | Free weights under Apache-2.0; you pay for the hardware | Lite $10 a month for 160 credits, Pro $25 for 500 credits, Pro Max $50 for 1,100 credits |
| Commercial use | Allowed under the Apache-2.0 license | Commercial use on every plan |
When Ovi is the better pick
If you already own a big GPU, want to run thousands of clips without a meter, or need to change the model itself, Ovi's open weights and Apache-2.0 license are hard to beat. Researchers and tinkerers should start there.
If you want the clip today, want to compare engines on the same line and keep going into captions, edits and whole films, we built Pixel Dojo for that.
What talking clips cost on Pixel Dojo
Plans are Lite $10 a month for 160 credits, Pro $25 for 500 credits, Pro Max $50 for 1,100 credits. Every plan includes every model and commercial use.
In our test the WAN 3.0 clip cost 10 credits, Veo 3.1 Fast 27, Kling 3.0 27.5 and Seedance 2.5 30. Each model page shows its price before you generate.
Under our refund policy, a first purchase can be refunded within 7 days, in full if under 20% of the credits are used and pro-rated after that.
Ovi questions
What is Ovi?
Ovi is an open source model from researchers at Character AI and Yale that generates video and matching audio together from text, or from text plus an image. Its README lists an Apache-2.0 license.
What do I need to run Ovi?
Ovi's README says the minimum is 32GB of GPU memory, or about 24GB with an fp8 version that has slight quality loss.
What is a good alternative to Ovi?
Pixel Dojo runs Veo 3.1, Seedance 2.5, Kling 3.0 and WAN 3.0 in the browser. All four write picture, speech and sound together, with each price shown before you generate.
Which engine is best for a talking clip with sound?
In our pier test Veo 3.1 Fast gave the cleanest spoken performance, WAN 3.0 showed the most of the sound cues on screen for the fewest credits, and Seedance 2.5 had the strongest mood.
Can I start from my own image like Ovi?
Yes. Make or upload a still, then give it to Kling 3.0, Seedance 2.5 or another engine as the first frame. The accordion clip on this page was made that way.
Video Effects New
No prompt? Try a one-click effect
283 ready-made effects, from AI Hug and Earth Zoom In to product spins and listing walkthroughs. Upload a photo and press Generate. No prompt to write.
Try Pixel Dojo as your Ovi alternative
Write a prompt above or start from any example on this page. One subscription covers every model on Pixel Dojo.
Start with Veo 3.1Sources
Facts about Ovi come from these public pages, checked on Oct 15, 2026. Prices and plans change; check the source before you buy.