---

title: 'How to Make a Tiny Desk AI Video (Step-by-Step Guide, 2026)' description: 'Learn how to make a Tiny Desk AI video from three photos: what to prepare, two ways to generate it, photo tips, a DIY prompt template and common mistakes.'

How to Make a Tiny Desk AI Video: A Step-by-Step Guide

The short version: A Tiny Desk AI video is a short, vertical, live-room band performance generated by AI. The easiest way to make one is to give a tool three portraits (a lead singer, a sax player and a keys player) and let it generate the room, the camera moves and the sound. You can also build one yourself with a general video model and a carefully written prompt. This guide covers both.

Short Answers

• What do I need? Three clear, front-facing portraits of three different people or characters, and permission to use them.
• How long is the video? The viral template is a vertical clip of roughly 19 seconds with five shots. Most generators produce clips of 10 to 15 seconds.
• Which format? 9:16 vertical MP4, with audio generated alongside the picture.
• What is the fastest route? A purpose-built tool such as TinyDeskAI, which turns three photos into a finished clip.
• Is it free? It depends on the tool. TinyDeskAI sells one-time credit packs, with no subscription and no watermark.
• Is it a real Tiny Desk concert? No. It is a parody. Nothing in it is made by NPR or the Tiny Desk Concerts series.

Step 1: Understand What You Are Recreating

The format comes from a small-room live performance: a band plays close to the camera around one desk, with shelves of records, books and knick-knacks behind them. The 2026 meme version keeps one famous performance structure and swaps the people in it. If you want the background first, read our guides to what Tiny Desk is and the Outstanding Tiny Desk trend.

The shot structure matters, because that is what makes the clip feel like the meme:

1. Wide shot. The lead stands at the mic, one hand on chest, while the keys player comps chords behind.
2. The big note. A tight close-up as the lead belts and holds one long note.
3. The sax solo. Focus pulls to the sax player over the lead's shoulder.
4. The wrap. Back to wide: a raised fist, a few claps and a small "woo".

Step 2: Prepare Your Three Photos

Photo quality decides most of the result. For each of the three roles (lead singer, sax player, keys player):

• Use one person per photo. Avoid group shots.
• Make sure the face is fully visible and looking at the camera.
• Use even lighting, with no heavy shadows, sunglasses or strong filters.
• Head and shoulders is enough.
• Use three different people or characters, so the band reads as a band.

AI-generated characters also work well, as long as each portrait is sharp and front-facing.

Step 3: Choose How to Generate It

Option A: Use a purpose-built tool (fastest)

We built TinyDeskAI for exactly this. The workflow takes three steps:

1. Add three photos. One for the lead singer, one for the sax player and one for the keys player.
2. Generate. Spend 20 credits. The set, the camera moves and the song are already written for you.
3. Download the clip. You get a vertical MP4 with original sound and no watermark.

A few practical details:

• You can choose between a new-song record-shop room and an "Outstanding" version built around the meme.
• Each video costs 20 credits. Credit packs start at $9.90 for 20 credits, and credits never expire. There is no subscription.
• If a render fails, its credits go back to your account automatically.
• A video usually takes a few minutes, and you can close the tab while it renders. Finished clips appear under "My videos".

It needs three portraits at the moment, so a single-photo version is not available yet. Likeness, lip-sync and finger movement vary from clip to clip, so a sharp, well-lit photo helps more than anything else.

Option B: Build it yourself with a general video model

If you prefer full control, you can use a video model that accepts reference images and generates native audio. Seedance-style models are a common choice. The approach:

1. Upload your three portraits as image references and label them by role.
2. Write a prompt with five parts: the format, the setting, the audio, a timed shot list, and a negative list.
3. Keep the prompt short enough for your tool's character limit, and tell the model to keep everything except the people unchanged.

Here is a compact template you can adapt:

Vertical 9:16, 10s, handheld live-session feel, shallow depth of field on close-ups, warm filmic grade. @Image1 = lead vocalist, @Image2 = saxophonist, @Image3 = keyboard player; keep identities consistent in every shot. Exactly three people, no audience.

SETTING: cozy record-shop rehearsal room, warm wooden shelves packed with vinyl records, hanging plants, amber lamps. Left: wooden desk with a red keyboard and mugs. Center: mic stand for the lead. Right: gold saxophone.

AUDIO: native live-room sound, soulful pop-groove at about 110 BPM. Warm electric-piano chords, a powerful emotional English lead vocal, a bluesy alto-sax solo entering at 6s, a few claps and a "woo" in the last second.

TIMELINE: 0-2.5s wide shot, lead sings softly with a hand on chest. 2.5-5s tight close-up, lead belts one long note. 5-6s big smile to camera. 6-8s rack focus to the sax player, who plays the solo. 8-10s wide again, lead raises a fist and waves, band applauds.

NEGATIVE: extra people, text, logos, watermarks, distorted hands, room layout changes.

Option B takes more trial and error, but it lets you change the room, the song and the shot list.

Step 4: Review and Fix Common Problems

• Faces drift between shots. Use sharper, more frontal photos and repeat the role labels in the prompt.
• Lip-sync looks off. Keep the shot list simple and avoid describing mouth movement in detail.
• Extra people appear. Add "exactly three people, no audience" and put extra performers in the negative list.
• The background changes between cuts. State that the room layout and props stay the same.

Step 5: Post Responsibly

Only use photos of adults who have given you permission, or fully fictional AI-generated characters. Never use a photo of a minor. Do not use the result to mislead anyone, to impersonate a real person, or in advertising or endorsements. Turn on your platform's AI label when it offers one. Tiny Desk is an NPR brand, and tools like ours are independent parody tools with no affiliation to NPR.

FAQ

Can I make a Tiny Desk AI video from one photo? Not with TinyDeskAI yet. It needs three portraits, one for each role.

Do I need the original performance footage? Not with a photo-based tool. With a general video model, an optional reference clip can help lock the shot rhythm.

Will the video look exactly like the person in the photo? It will look like them rather than be a perfect copy. Results vary from clip to clip.

Is there a watermark? Not on TinyDeskAI videos.

Where can I start? Go to aitinydeskvideo.org, add your three photos and generate your first clip.

---

TinyDeskAI is an independent, unofficial parody tool and is not affiliated with, endorsed by or sponsored by NPR or the Tiny Desk Concerts series. Information current as of October 28, 2026.