Runway Image to Video: Gen-4.5 Steps, Prompts and Credit Costs

Runway turns a still photo into video by using your image as the first frame and letting Gen-4.5, its current flagship video model, animate what happens next. In November 2026 that means clips of 2 to 10 seconds at 720p, made in the desktop web app, with no sound of their own. Each second costs 12 credits, and the feature needs the Standard plan ($15 a month, or $12 a month billed yearly) or higher. One rule decides most results: describe the motion, not the picture. The photo already shows the scene. Your prompt only tells the camera and the subject what to do.

That makes Runway a strong fit for short product shots, animated renders and social b-roll. It is a weak fit for anything that must run longer than ten seconds in one take or ship with synced sound straight out of the generator. Other models on the same plan close part of that gap. Gen-4.5 does not.

What Gen-4.5 does with your photo

Gen-4.5 launched in December 2025 and gained first-frame image to video in January 2026. The uploaded image is locked in as frame one; everything after it is generated. Output is 720p at 24 or 25 fps, and the help page lists 1280x720 for a 16:9 image-to-video clip ^[1]. You choose a duration between 2 and 10 seconds, and Runway notes that prompts asking for several actions in a row benefit from the longer end. A finished clip can be upscaled to 4K from the option under the generation.

Gen-4.5 runs in the web app. It is not in the iPhone or Android apps, though the Premiere Pro and After Effects plugins, the DaVinci Resolve Studio plugin, Runway MCP and the API (billed in separate API credits) reach it too. The Free plan's 125 one-time credits do not unlock it, and Runway does not publish the exact list of models Free accounts get. There is no audio track. Add voice, effects or music afterwards with Runway's audio tools, or pick a model that generates sound, covered below.

The older Gen-4 and Gen-4 Turbo still work from an image. Both require one as the first frame, make 5 or 10 second clips at 24 fps, and Runway labels them an older generation. Turbo costs 5 credits a second instead of 12.

Prepare the image before you upload

Most failed clips start with the source file. Runway's realism guide asks for reference photos of at least 1080p ^[2], because blur and low resolution carry straight into the video. Run through these checks first:

• Resolution: 1080p or more, free of compression artifacts. A weak photo can go through the Upscale Image app first (up to 48 credits per image).
• File type: JPEG, PNG or WebP. GIF is not accepted.
• Motion cues: motion blur, dust clouds or a blurred background tell the model things are already moving. Remove them if you want a calm shot.
• Framing: crop to the format you will publish. Runway suggests 9:16 for TikTok, Reels and Shorts, 16:9 for YouTube and LinkedIn, and 1:1 for an Instagram feed post.

If you have no photo yet, Gen-4 Image can create the still inside the same workspace for 8 credits at 1080p, then hand it to Gen-4.5. Two rights points matter when you upload real photos. You must hold the rights to the people, products and places in them. And Runway's consumer Terms let it train its models on your inputs and outputs; a no-training promise comes only with an Enterprise contract.

Write a prompt that moves the picture

Runway's image-to-video guide, written for Gen-4.5, suggests one structure: the camera does a movement as the subject does an action, followed by any extra detail ^[3]. The detail can cover the environment, style, timing, direction and speed. Re-describing what the image already shows wastes words and can pull the model off course. Runway's photo-to-video tips put the sweet spot at about 20 to 40 words built from camera movement, scene action and atmosphere.

Three habits come straight from Runway's guides. Ask for one clear motion per clip; one camera move gives the most stable result on most models. Phrase everything positively, because a negation such as no camera shake can produce exactly that shake. And write prompts in English when you can; other languages work, but less consistently. Here are three prompts in that shape, one per common job:

• Product: The camera pushes in slowly as steam rises from the coffee cup and morning light moves across the table.
• Portrait: The camera is entirely motionless as the woman turns toward the window and smiles, her hair lifting in a light breeze.
• Property: The camera glides forward through the open doorway as the curtains sway and sunlight spreads across the wooden floor.

A truly still camera is hard to get. Runway's advice is to describe the movement inside the scene and say plainly that the camera stays motionless; the Animate Frames app is the alternative for locked-off shots. Unwanted cuts usually mean the image and prompt suggest two scenes. A longer duration helps, and so does asking for one continuous, uncut shot. For a sequence, rough timestamps such as [00:01] followed by the event guide timing only loosely. Want a fuller draft? The star icon runs Enhance Prompt; JSON prompts, Runway says, are no more accurate than plain sentences.

When one frame is not enough

Gen-4.5 takes a first frame only. To pin down the ending too, Runway offers the Animate Keyframes app, which builds a smooth transition between a start and an end keyframe. Runway's plan also unlocks other labs' video models inside the same workspace, all from Standard up, and several accept more frame control:

The difference that matters most is length and sound. A thirty-second shot or a clip that needs ambient audio from the start points away from Gen-4.5. A ten-second silent move that will get music in the edit does not. MiniMax H3 and Wan 3.0 also take a first frame plus an optional last one, as a separate mode from reference images. Kling 3.0 Motion Control animates a still image with the movement from a reference video. Our model comparison sets them side by side on resolution and inputs. Kling 4.0 on Runway also has a 25% credit promotion that ends November 30, 2026.

Chain short clips into a longer shot

Runway states plainly that no model makes a long-form video of ten minutes or more in one pass, and it says most image-to-video tools handle clips of 5 to 10 seconds well. The last-frame method fills the gap. Take the final frame of a finished clip, use it as the image input for the next generation, and join the clips in a video editor. In Workflows, the Last Frame node pulls that frame out for you, and the Stitch Video app combines clips into one file of up to 60 minutes.

Each new link starts from a frame the model drew, not your original photo. Check that frame for soft faces or warped edges before you reuse it, because the next clip inherits them. Write each prompt as its own single scene. A long sequence is several short decisions, not one big one. Runway says the architecture firm KPF now animates static renderings in house in a few hours, work it previously outsourced. That job suits the method well. One render. One camera move. One short clip.

What each clip costs

The rates are fixed per second: Gen-4.5 at 12 credits, Gen-4 Turbo at 5, and Runway shows the full cost before you press generate ^[4]. By our arithmetic, the 625 monthly credits on Standard cover about 52 seconds of Gen-4.5, or ten five-second clips. Pro's 2,250 credits cover about 187 seconds, or 37 clips. The same Standard allowance stretches to 125 seconds on Gen-4 Turbo. Kling 4.0 at 1080p gets five five-second clips out of it.

Extra credits on paid plans come in packs of 1,000 for $10 and never expire. At that rate a five-second clip at 12 credits a second costs $0.60 and a ten-second clip $1.20. Exporting ProRes or a PNG sequence is a Max feature and lifts Gen-4.5 to 17 credits a second.

Credits come back only when a generation ends in an error. A clip you dislike keeps its cost, so your real price per usable clip runs above these figures; Runway itself says to judge cost per usable output. Plan credits on Standard and Pro reset each billing date, and refunds are discretionary; our Runway pricing guide covers rollover, refunds and yearly billing.

The bottom line

Image to video is where Runway is easiest to control: your photo fixes the look, and a short motion prompt does the rest. Start from a sharp image of at least 1080p, cropped to its final format. Write one camera move and one action, phrased positively. Accept the limits of Gen-4.5 in November 2026: 720p, ten seconds and silence. Reach for Kling 4.0 or Seedance 2.5 on the same plan when you need longer shots, set endings or built-in sound. Budget per usable clip, not per generation, and Standard is enough to find out whether the workflow suits you.