How to Make an AI Fight Video from Your Photos (VdoBloom AI Fight Tab)
Turn one or two photos into a martial arts or street fight clip. Real templates, real credit costs from 10 credits, prompt tips and when to switch models.
To make an AI fight video, open the AI Fight tab on VdoBloom, upload one or two clear photos, pick the Martial Arts or Street Fight template, and generate — a 5-second 720p clip on the default Runway Gen-3 model costs 10 credits, which is exactly the free signup balance and about $0.50 at Lite plan rates. The AI Fight effect is a photo-motion tool: it takes a still photo of a person (or two photos of two different people) and animates them into a choreographed action scene. No green screen, no stunt double, no video footage required.
What the AI Fight effect actually does — and what it does not
The tab sends your photo plus an action prompt to a video model, which re-animates the subject. What you reliably get is 5 to 10 seconds of combat motion: stances, strikes, camera push-ins, impact lighting. What you do not get is frame-perfect identity retention. Faces drift on fast motion, hands break first, and the further the model moves the body from its pose in the source photo, the less the person looks like themselves.
That is why photo count is a toggle. In single-photo mode the effect builds one fighter and invents an opponent; in two-photo mode it takes Fighter 1 and Fighter 2 as separate uploads and puts both in the same scene — the more impressive result, and the harder one to get right.
One rule that is not optional: if the photo is of a real person other than yourself, you must have that person’s consent before animating them into a fight video.
The photo that works best
- Full body or three-quarter body. A head-and-shoulders selfie gives the model no legs, arms or stance to work with, so it hallucinates them. Waist-up is the minimum.
- Standing, arms visible, not crossed. Arms tucked behind the back or out of frame produce the worst limb artefacts.
- Even lighting, no heavy shadow across the face. The model re-lights the scene dramatically; it cannot recover facial detail that was never in the source.
- For two-photo mode, match the framing. Two photos shot at roughly the same distance and angle composite far better than one close-up plus one full-length shot.
The two templates
The tab ships with two preset templates, and each one carries a different prompt depending on whether you uploaded one photo or two. You do not have to write anything — selecting a template fills the prompt box for you.
| Template | Look | Prompt it loads (two-photo mode) |
|---|---|---|
| Martial Arts | Epic martial arts combat scene | two people engaged in an epic martial arts fight scene, dynamic combat moves, cinematic quality |
| Street Fight | Intense street fight action, harder lighting | two people in an intense street fight action scene, powerful moves, dramatic lighting, cinematic quality |
The default prompt before you touch a template is: two people in an epic action fight scene, dynamic martial arts moves, cinematic quality. Martial Arts gives cleaner, more readable choreography; Street Fight gives grittier contrast and a handheld feel that hides limb artefacts better, which makes it the safer choice on a weaker source photo.
What it costs
The AI Fight tab opens on Runway Gen-3, and that is deliberate: at 10 credits for 720p and 5 seconds it is the only effect model a brand-new account can afford on its free signup balance. Every price below is a real number from the credit table, converted at the Lite plan rate of $15 for 300 credits, so 1 credit is $0.05.
| Model | Tier | Credits | Cost | Best for |
|---|---|---|---|---|
| Runway Gen-3 (default) | 720p / 5s | 10 | $0.50 | First test, and free on signup credits |
| Runway Gen-3 | 1080p / 5s | 20 | $1.00 | The keeper version of a shot you already like |
| Runway Gen-3 | 720p / 10s | 20 | $1.00 | A longer exchange rather than a single strike |
| Runway Gen-3 | 1080p / 10s | 30 | $1.50 | Full-length final cut |
| Vidu Q2 Turbo | 720p / 5s | 18 | $0.90 | Cheap second opinion on the same photo |
| PixVerse v4.5 | 720p / 5s | 22 | $1.10 | Stylised fights — anime, comic, cyberpunk styles |
| PixVerse v4.5 | 1080p / 8s | 50 | $2.50 | Longest stylised option on this tab |
| LTX-2 Fast | 1080p / 6s | 20 | $1.00 | 1080p at a 720p price |
| Kling 2.5 Turbo | 5s | 40 | $2.00 | Best physics and body coherence on fast motion |
| Kling 2.5 Turbo | 10s | 80 | $4.00 | Long take where the choreography must hold up |
New accounts get 10 free credits with no card, which is exactly one 720p Runway Gen-3 fight clip. Paid plans start at $15 per month for 300 credits and remove the watermark with commercial rights; one-time credit packs start at $2.49 for 75 credits and never expire. Full detail is on the pricing page.
The exact steps
- Go to the AI Fight tab.
- Choose one photo or two. With two, the left slot is Fighter 1 and the right slot is Fighter 2 — the order shows up in the composition, so put your hero on the left.
- Upload. Waist-up or fuller, arms visible, clean background.
- Pick Martial Arts or Street Fight. The prompt box fills in automatically; edit it if you want a specific setting, like a rooftop at night or a dojo.
- Set duration and quality. Start at 720p and 5 seconds — 10 credits — before you spend on 1080p.
- Pick an aspect ratio. Runway Gen-3 offers 16:9, 4:3, 1:1, 3:4 and 9:16. Use 9:16 for Reels, Shorts and TikTok.
- Generate, then review at full size before you re-run. Most fixes are prompt fixes, not model fixes.
How to prompt a fight scene well
The template prompts are a starting frame, not a ceiling. Three additions do most of the work:
- Name one move, not a sequence. A spinning back kick lands far better than an exchange of punches and kicks. Five seconds only holds one beat.
- Name the environment. Rain-soaked alley, neon signs overhead beats dramatic lighting, because it gives the model something concrete to light.
- Name the camera. Slow-motion low angle tracking shot or handheld camera pushing in changes the result more than any adjective about intensity.
What to avoid: blood, weapons aimed at a face, and anything that reads as real violence against a real identifiable person. VdoBloom’s flexible content tier covers action, dance, fitness and fashion that stricter tools refuse, but explicit and illegal content stays blocked, and a graphic gore prompt will simply fail.
A worked example
Two friends, two waist-up phone photos taken in the same room. Fighter 1 in the left slot, Fighter 2 in the right. Template: Street Fight. Prompt edited to: two people in an intense street fight in a rain-soaked alley at night, one throws a spinning back kick, neon signs reflecting off wet ground, handheld camera pushing in, slow motion, cinematic quality. Settings: Runway Gen-3, 720p, 5 seconds, 9:16. Cost: 10 credits, $0.50. Two attempts later the keeper was re-run at 1080p for 20 credits. Total spend for a finished vertical clip: 40 credits, $2.00.
When to pick a different model
Stay on Runway Gen-3 while you are iterating; it is the cheapest way to find out whether the photo works at all. Switch to Kling 2.5 Turbo when the motion is the point and Gen-3 keeps bending a limb — it holds body coherence through fast movement better than anything else on this tab, and 40 credits for the shot that actually works beats four failed 10-credit attempts. Switch to PixVerse v4.5 when you want the fight stylised rather than photoreal; it exposes anime, comic, 3D animation, clay and cyberpunk styles plus explicit camera-movement presets, which is a different creative job entirely.
Frequently asked questions
How many credits is an AI fight video?
10 credits for 720p and 5 seconds on the default Runway Gen-3 model, 20 for 1080p at 5 seconds or 720p at 10 seconds, and 30 for 1080p at 10 seconds. Kling 2.5 Turbo is 40 credits for 5 seconds and 80 for 10. At Lite plan rates, 1 credit is $0.05.
Can I put two different people in the same fight video?
Yes. Toggle the tab to two photos and upload Fighter 1 and Fighter 2 separately. The model composites both subjects into one scene, and each template carries a distinct two-photo prompt. You need consent from anyone in those photos who is not you.
How long can the video be?
5 or 10 seconds on Runway Gen-3 and Kling 2.5 Turbo, 5 or 8 seconds on PixVerse v4.5, and 6, 8 or 10 seconds on LTX-2 Fast. There is no 30-second option on this effect; longer action is built by generating several clips and cutting them together.
Why does the face change during the fight?
Fast motion is the hardest case for identity retention. Use a sharper, better-lit photo, name a single move instead of a sequence, shorten to 5 seconds, and try Kling 2.5 Turbo, which drifts least on rapid movement.
Ready to try it?
Create your first AI video in minutes — no credit card required.
Start Creating Free →