Which photos work
Hit the Road Dude Video From a Photo
A Hit the Road Dude video from a photo is a 10-second vertical clip, 9:16 at 720p, in which the person in your photo is the driver of the white pickup, with sound. You upload one photo of a face and wait about three minutes. The clip is AI-generated, and this page does not pretend otherwise.
What decides the result most is the photo. This page is not about how to press the buttons. It is about which kinds of photo give you a good clip, which give you a strange one, and why.
Which photos work, and what to expect
Six kinds of photo, with the result we expect and the reason. These are expectations, not promises. Models vary from one run to the next.
| Photo type | What you will likely get | Why |
|---|---|---|
| Front-facing selfie, good light | The best case. A clear face as the driver, the right mustache or none, and the expression carried through | The model can see both eyes, the mouth and the jaw, so it has nothing to guess |
| Glasses (clear lenses) | Usually fine. The frames carry over and the face stays recognisable | Clear lenses leave the eyes visible. Thick frames can shift a little between frames |
| Sunglasses | Risky. The lenses can melt, change shape or become part of the face | The model cannot see the eyes, so it fills the space with something plausible |
| Side profile | Often drifts. The face can turn to look like somebody else halfway through | Half the face is hidden, so the model invents the other half and the invention moves |
| Group photo | Unpredictable. It may pick the wrong person as the driver, or blend two faces | There is no single face to follow, so it has to choose |
| Low light | Grainy or flat. Skin can look smooth and eyes dark, and details are lost | Little information in the pixels, so the model fills in with an average |
| A pet | It can work and it is often funnier than the human version. Expect odd fur and an odd mouth | The model was trained mostly on people, so it reaches for human features and guesses the rest |
One face, clearly visible
The rule that explains most of the table is simple. The model needs one face it can see. A front-facing photo, even a plain one, gives it that. A cropped passport style picture will do better than a lovely wide shot of a party.
Do not worry about being photogenic. A tired face in a kitchen works fine. What matters is that nothing covers it and nothing else competes with it for attention.
If you have several photos, pick the one where you are looking close to the camera with a neutral face. The model then adds the mouth movement for the line.
What hurts a photo
Four things make a result worse. Covering part of the face: sunglasses, a hand, a mask, a large hat. Turning away: a side profile or a head tilted far. Crowding: more than one person. And darkness: low light with heavy grain.
None of these is fatal. We have seen odd results turn out well. But if you want the best odds, avoid them. If you only have a poor photo, the clip may still be funny, and a failed render costs nothing.
Heavy filters are a fifth. A beauty filter or a cartoon effect changes the face before the model sees it, and the result keeps those changes.
What about the clothes and the truck
The shot is a driver in a white pickup, seen through the driver window, so the clothes are mostly hidden by the door. The model decides what you are wearing from the photo and the scene. A collared shirt in the photo tends to carry across.
You are not able to choose the truck or the road. The scene is fixed. The one thing that comes from you is the face.
The voice and the sound
The clip has sound and two voices: the driver, who is you on screen, and a visitor off camera who answers. Both are generated together with the video. Neither is your voice. We do not clone anyone’s voice from a photo, and a photo carries none.
If the voice is wrong for the face, that is expected. It is a character voice, and the visitor’s voice is a different one on purpose. People often find a mismatch funny, but you should know about it before you pay. Sometimes the mouth moves on the visitor’s line; that is the model, and it is the most common flaw.
Whose face goes in
Use your own face, or the face of someone who has said yes. The funniest clips are of friends and family who are in on it. Do not put a private person into a video they would not want to be in.
We also do not use the face of public figures and ask you not to. This is a joke for ordinary people. And do not present the result as a real recording. It is AI-generated, and it looks it if you watch closely.
What it costs
One video is 13 credits. A three video pack is $19.90 and a five video pack is $29.90. There is no subscription and no free credits. A failed render costs nothing, so a bad photo costs you time and not money.
If you are not sure a photo will work, try the best one you have first. The table above tells you what to expect from the others.
Size, crop and sharpness
A photo does not need to be large. It needs to be sharp where it counts. A small picture where the face fills the frame beats a huge picture where the face is a few dozen pixels across at the back of a room.
Crop in if you can. Leave a little space above the hair and below the chin, and nothing else. A face that touches the edge of the picture gives the model less to work with around the jaw and ears.
Avoid screenshots of other videos, because they carry compression blocks that the model treats as part of the face. A photo straight from a phone camera is the safest source.
If you are unsure between two photos, use the one with the more even light. Hard shadows across half a face behave a lot like a side profile.
Asked in the comments
Video from a photo: questions
Can I make the video from just one photo?
Yes. One photo of a face is all that is needed. A clear, front-facing photo in good light gives the best odds.
Will a selfie work?
A front-facing selfie in good light is the best case. Low light and heavy filters make it worse.
What if my photo is a side profile or a group?
A side profile can drift, because half the face is hidden. A group photo makes the model choose a person, and it may choose the wrong one. Use a single front-facing photo if you can.
Can I use my pet?
You can. It often works and it is often funny. Expect odd fur and an odd mouth, because the model leans on human features.
How long does it take and how long is the clip?
About three minutes. The clip is a 10-second, 9:16 at 720p, with sound.
Is it real?
No. It is AI-generated, and you should not present it as a real recording.