Closed on Sunday AI Text to Video: No Photo Needed

You can make a Closed on Sunday AI video from text alone: type a description of the scene, pick a preset, and the tool generates a 12-second silent clip without any photo upload. Text to video is the fastest route into the trend, and it works best when you care about the mood of the shot rather than a specific person's face. This guide covers what to write, what the model fills in on its own, and where a photo still gives you a better result.

Text to video suits the version of this trend where the scene matters more than the person. Use it when:

• You have no usable photo, or the photos you have are low resolution, blurry, or badly lit.
• You want a wide shot of a choir on a hillside rather than a close-up of one face.
• You are testing ideas quickly and do not want to prepare an image for every attempt.
• You want a generic lead figure and are not trying to depict yourself or anyone specific.

Skip it when the whole point is a recognisable person. If you want your own face, your friend's face, or a family member in the frame, start from a photo instead. Text to video cannot reproduce a real likeness, and asking it to will only waste credits.

If you are still deciding whether to try the trend at all, the overview of what the trend is is a sensible first read. The original music video came out on 28 November 2019; the AI remake wave took off around 14 October 2026, and most of the clips you have seen are built either from a photo or from a written prompt.

When you type a prompt, you are not directing every element. You are setting a few constraints and letting the model resolve the rest. Here is what it decides on its own unless you specify otherwise.

The lead. A single central figure, usually framed from the waist up or in a medium shot. Age, build, hair and clothing are inferred from whatever you wrote. If you wrote nothing about appearance, expect a generic adult in neutral clothing.

The choir. A group behind or beside the lead, typically in rows. The model chooses how many people, how they are spaced, and whether they sway, clap, or stand still. Choir size is loosely controlled by words like "small group" or "large choir", but the exact count is not something you can pin down.

The light. This is where the trend's look lives. Left alone, the model tends toward warm backlight, a bright sky, and soft haze. Words like "golden hour", "overcast", "harsh midday" or "backlit" shift it noticeably.

The camera. Expect a slow push-in or a gentle drift by default. You can ask for a static shot or a slow pan, and it will usually comply, but very specific camera moves are hit and miss in a 12-second clip.

Long prompts tend to produce muddled clips. Four short lines, one per element, gives the model enough to work with and keeps you able to tell which line caused a problem.

Wide shot of a choir standing on a dry grassy hillside at golden hour.
Lead singer in the centre, dark coat, short hair, calm expression.
Choir in loose rows behind, swaying gently, warm backlight and haze.
Slow push-in, static horizon, no text.

A few practical notes on this structure:

• Put the shot type first. "Wide shot" or "medium shot" changes the composition more than any other single word.
• Keep the lead description to three or four attributes. More than that and the model starts dropping them.
• Describe motion for the choir, not the lead. A still lead with a gently moving choir reads better than both moving at once.
• End with what you do not want. "No text" and "no logos" are worth including.

For more on prompt structure across the tool, see the prompt guide.

You cannot get a specific real person from text, but you can get a consistent type of person. The attributes that actually register:

• Age range. "In his thirties", "middle-aged", "young adult". Vague ranges work better than exact numbers.
• Clothing. "Dark long coat", "plain white shirt", "oversized knitwear". Colour and silhouette matter more than fabric.
• Hair. "Short dark hair", "close-cropped", "shoulder-length and loose". Avoid brand names or very specific styles.
• Build and posture. "Tall and still", "slight build, hands at sides". Posture affects the mood more than body type.
• Expression. "Calm", "eyes closed", "looking slightly off camera". This is the single biggest lever on tone.

Leave out anything you do not care about. Every unnecessary detail is another thing the model can get wrong, and a wrong detail in a 12-second clip is hard to ignore.

The setting does a lot of the work in this trend, and it is easier to control than faces.

For the hillside:

• Ground. "Dry grass", "dusty earth", "low green scrub". Dry grass and dust read as the classic look.
• Horizon. "Low horizon, wide sky" gives you space above the choir. "Hills behind" closes the frame in.
• Time of day. "Golden hour" for warmth, "late afternoon" for softer contrast, "overcast" for flat and moody.
• Weather. "Light haze" adds depth. "Clear sky" is cleaner but flatter.

For the choir:

• Size. "Small group of about a dozen" or "large choir filling the frame". Treat the number as a hint, not a setting.
• Arrangement. "Loose rows", "clustered", "standing in a line".
• Motion. "Swaying gently", "clapping slowly", "standing still". Slow motion suits a 12-second clip better than anything fast.
• Framing relative to the lead. "Choir behind the lead, slightly out of focus" keeps attention where you want it.

If you want to branch out from the standard hillside, the choir ideas page has alternatives worth trying.

Be realistic about what text to video will not do.

• No likeness. You cannot generate a recognisable real person from a description. Names of public figures will not produce them, and attempting it is not what this tool is for.
• Hands are unreliable. Clapping, raised hands and clasped hands are the most common source of visual errors. If hands are not essential, frame them out or keep them at the sides.
• Crowd faces blur or distort. In a large choir, individual faces in the back rows often go soft or wrong. Wide shots hide this; close shots expose it.
• Text and signage fail. Do not ask for words in the frame.
• Continuity across clips is manual. Each clip is generated fresh. If you need two shots to match, describe them identically and expect small differences.

The most reliable setup for most people is a hybrid: a photo for the lead, and typed text for everything else. You get a face you actually recognise, and you keep control of the setting without needing to source a background image.

In practice:

• Upload a clear, front-facing photo with even lighting for the lead.
• Type the location, light and choir description as you would in a pure text prompt.
• Keep the typed portion short. The photo already carries the lead; the text only needs to place them.

This avoids the two most common failure modes at once: a generic lead you do not care about, and a photo whose background fights the scene you wanted. If you are choosing an image for the lead, the photo selection guide covers what tends to work.

Our clips are 12-second silent MP4s. Pricing is per clip and depends on resolution:

• 480p, 12 seconds: 15 credits
• 720p, 12 seconds: 30 credits
• 1080p, 12 seconds: 60 credits

The smallest one-time credit pack is $9.90 for 15 credits, which is one 480p clip. Failed renders are refunded automatically, so a bad generation does not cost you.

Practical advice:

• Draft at 480p. Text prompts need iteration. Get the composition and mood right at the lowest tier, then regenerate the winner at 720p or 1080p.
• 720p is the sensible default for social posting. 1080p is worth it if you plan to crop or zoom in an edit.
• One lead per clip. Do not try to fit two central figures into 12 seconds.
• Silent by design. Add your own audio in your editor; the clip arrives without a soundtrack.

There are eight presets plus a typed choir option, so you can start from a look and adjust rather than writing everything from scratch.

Closed on Sunday AI is an independent tool. It is not affiliated with Ye, Sunday Service, Jesus Is King or Chick-fil-A.

No. Text to video generates the whole scene from a written prompt. A photo is only necessary when you want a specific, recognisable person as the lead.

No. Text prompts produce a generic figure based on the attributes you describe. For a real likeness you need to start from a photo of that person.

Every clip is 12 seconds and silent. It is delivered as an MP4, and you add music or voiceover yourself in an editor.

15 credits at 480p, 30 at 720p and 60 at 1080p, all for 12 seconds. The smallest one-time pack is $9.90 for 15 credits. Failed renders are refunded automatically.

Hands and small faces in a crowd are the hardest things for the model to render consistently. Frame them out, keep hands still, or use wider shots where individual faces are not the focus.

Yes. Email delivery sends a private link that stays live for seven days, and the recipient does not need an account to open it.