How to make an architecture video from Blender renders: 9:16 output settings or six stills

You modeled the project in Blender, the stills look right, and now you need a short vertical video for a client or for Instagram. Blender can render the video itself, but the settings that matter are spread across three panels, the render engine changes the time per frame by a wide margin, and rendering straight to a movie file is the mistake the manual itself warns against. This guide covers the two routes, a camera animation rendered in Blender or six stills sent to an image-to-video tool, with the exact Output Properties settings for a 1080 × 1920 file and the arithmetic that tells you which route your deadline allows.

Key takeaways

• For a vertical film, set Resolution X to 1080 and Y to 1920 in the Format panel; right-clicking the resolution fields and choosing Swap Dimensions flips an existing landscape setup.
• Render the animation as a PNG image sequence, then encode it to an H.264 MP4 in Blender's Video Sequencer; the Blender manual calls rendering directly to a movie file "highly not recommended."
• EEVEE renders far faster than Cycles because it rasterizes instead of path tracing, but the manual is explicit that "Cycles will always provide more physically accurate renders."
• A 30-second film at 30 fps is 900 frames, so time one final frame and multiply before you promise a date.
• Six Blender stills sent to an image-to-video service avoid the frame budget, but the motion is an AI interpretation of the images rather than a path through your model.

Two ways to make an architecture video from Blender renders

To make an architecture video from Blender renders, either animate the camera, set the output to 1080 × 1920 and render the frames as a PNG sequence that you then encode into an MP4 in Blender's Video Sequencer, or render six finished stills and upload them to an image-to-video tool that returns a finished film. The first route is exact and costs hundreds of frames of render time; the second costs six renders and gives you an AI interpretation of them.

Cycles or EEVEE for an architecture animation

The engine you pick sets the time per frame, and the time per frame sets everything else. The Cycles page describes it as "Blender's physically-based path tracer for production rendering." The EEVEE page calls EEVEE "Blender's realtime render engine focused on speed and interactivity" and adds that it "is based on rasterization and is not a path tracer," that "Cycles will always provide more physically accurate renders," and that EEVEE therefore has a set of documented limitations.

Two Cycles settings help an animation stay on budget. In Render Properties > Sampling, Time Limit "stops rendering if the time exceeds the limit, even if the desired sample count hasn't been reached yet," which caps the cost of every frame. Noise Threshold turns on adaptive sampling; the manual says "typical values are in the range from 0.1 to 0.001, with lower values meaning better quality but longer render times" (Sampling, all Blender manual pages read as of October 2026).

Exact Blender output settings for a 9:16 video

All of these live in the Properties editor under Output Properties. The labels below follow the Blender 5.x manual.

Format panel

Format sets the frame. Resolution X and Y are "the number of pixels horizontally and vertically in the image." The manual's tip saves a step: "the image can be quickly switched between landscape and portrait orientation by right-clicking the resolution fields and choosing Swap Dimensions." The Percentage slider renders smaller test frames "that have the same proportions as the final image," so a 50% test of a 1080 × 1920 setup renders at 540 × 960. Leave Aspect X and Y at 1; the manual explains they exist for non-square TV pixels.

Frame Range panel

Frame Range sets Frame Start and End. Divide frames by frame rate for the running time; the manual's own example is that "a 250 frame animation at a frame rate of 30 will last 8.3 seconds." For a 30-second film at 30 fps, set Start 1 and End 900. Step "controls the number of frames to advance for each frame rendered," useful for a quick flip-book test.

Output panel

Output sets where frames go and in what format. When rendering an animation, "the frame number is appended at the end of the file name with four padded zeros (e.g. image0001.png)." In Blender 5.x, choose Media Type: Image and File Format: PNG for the frames; Media Type: Video "encodes each frame into a video container" and is what you use later, for the final encode. In 4.x versions, the video formats sit in the File Format list instead.

Render frames first, then encode the MP4

The manual's Rendering Animations page describes two approaches. The Direct Approach, rendering straight to a movie file, "is highly not recommended and not a standard practice." The Frame Sequence approach "is a much more stable approach," and the manual's rule is simple: "if the total render time is an hour or more, you want to use the 'Frame Sequence' approach." Its hints add that unless an animation renders in a few minutes, "it is best to render the animation as separate image files," using a lossless format such as PNG.

The arithmetic is the same one the manual uses: render a typical frame and multiply. Its example is a one-minute film at 24 fps, 1,440 frames, at 30 seconds a frame: 720 minutes, or 12 hours. For a 30-second vertical film at 30 fps, 900 frames at 2 minutes each in Cycles is 30 hours; the same 900 frames at 10 seconds each in EEVEE is two and a half hours. Time your own frame; those inputs are only examples.

To turn the frames into a video:

1. Render the animation (Render > Render Animation) to the PNG sequence.
2. Open the Video Sequencer and add an image strip with all the frames from the output folder.
3. Scrub through to check every frame is there, and add a music strip if you have a licensed track.
4. In Output Properties, switch Media Type to Video, then in the Encoding panel choose the container and codec, for example MPEG-4 and H.264.
5. Set Output Quality, or a custom CRF: "a smaller CRF results in better video quality but larger file size," on a scale of 0 to 51 for H.264.
6. Render Animation again; this pass only encodes, so it's quick.

If a second computer is free, the Output panel's "Primitive Render Farm" hint applies: point both machines at a shared folder, "disable Overwrite, enable Placeholders," and start them both rendering.

Six stills to an AI film instead

If 30 hours of rendering doesn't fit the week, render six stills instead. Press F12 for each camera; the manual notes that still renders are not saved automatically and "must be saved manually from the Image Editor," so use Image > Save As and pick PNG or JPG.

Arch2Video takes exactly six images from one project (JPG, PNG, HEIC, TIFF or WebP, up to 30 MB each; the long side at least 512 px, and 1,500 px or more gives the sharpest scenes) and returns a 1080 × 1920 vertical H.264 MP4 of about 30 seconds: six connected five-second AI-generated scenes with an original soundtrack and an optional studio logo. The AI director plans the scene order, so it may not follow your upload order. Generation usually takes an hour or more, and you're emailed when it's ready. It costs USD 15 per video, a one-time card payment with no subscription; uploading is free and the price appears after the upload. There is no free trial and no free video.

It works from images, not from your .blend file, so there's no measured walkthrough, no prompt box and no editing timeline. Every film includes a small cast of original, fictional adults in at least one scene; people in your renders are not animated. Treat the motion as an interpretation, review it before posting, and present unbuilt work as concept work.

Pick the six as a set: context, approach, main space, secondary space, detail and a closing view, as in the six roles a project video needs. If the stills will also be posted as images, check rendering export settings for Instagram for color and size.

Which route for which audience

Design review and approvals. Render the animation. People will stop the film and compare it with the drawings, and a camera path through the model is the only version that holds up. Cycles is worth its time here if the film carries interior light.

Massing studies and internal reviews. EEVEE animation, often at 50% resolution. Speed matters more than glass.

Instagram, LinkedIn, website. Six stills to an AI film, or an EEVEE animation if you want full control. Nobody is measuring a social post, and a finished 30-second film in an hour beats a weekend of rendering.

For the full cost and accuracy comparison, see architectural animation vs. AI video. If part of your team renders in 3ds Max, the Corona and V-Ray version of this guide covers flicker settings for those engines.

Frequently asked questions

What resolution and frame rate should I set in Blender for a vertical Reel?

Set Resolution X to 1080, Y to 1920 and Percentage to 100% in the Format panel, and choose a frame rate of 24, 25 or 30 and keep it for the whole project. If your scene is already set to 1920 × 1080, right-click either resolution field and choose Swap Dimensions. Recompose the camera after swapping, because a landscape framing rarely survives a vertical crop.

Is EEVEE good enough for an architecture animation?

Often, for exteriors, massing studies and drafts, and it renders far faster. The Blender manual is clear that EEVEE "is based on rasterization and is not a path tracer" and that "Cycles will always provide more physically accurate renders," so interiors with glass, bounce light and reflections usually look better in Cycles. Render the same 3-second range in both and compare before committing hundreds of frames.

Why render PNG frames instead of an MP4 directly?

Because a crash, a power cut or a needed fix only costs you the affected frames. The manual calls the direct-to-movie approach "highly not recommended" and recommends the frame sequence for any render over an hour. You also encode once, at the end, in the Video Sequencer with full control over the codec and quality.

Can I use Blender stills with people in them for an AI film?

Yes, but the people won't be animated. Arch2Video replaces people in your images with its own small cast of original, fictional adults, and every film includes that cast in at least one scene. If you've modeled specific figures you want seen in motion, render the Blender animation instead; if the people are just scale figures, you can leave them in or take them out of the stills.