Subscribe
TEST RECORD · ← all field reports

FIELD REPORTS/VIDEO/ISSUE #133

Turn a photo and a shot by shot script into a video with dialogue, sound and music

Most video models give you one shot and leave the rest to luck. This one, from the team behind the Pika video app, lets you write the scene like a director, cut by cut, and the sound comes with the picture.

MODELPika-H3-Flex
PUBLISHEDOctober 30, 2026
READ TIME3 min
CATEGORYVIDEO

Field notes

01 What it does

You start from a photo or just a description, then write the scene as a list of shots: who is where, what happens, how the camera moves and when it cuts. Add a line of dialogue for up to four speakers, a background soundscape and a music style. You get back a clip of 5 to 15 seconds, and the audio is made together with the picture, so the ball cracks when it is hit and the voice comes from the person who is talking.

You can also hand it references to keep things consistent: up to nine images, three short videos and three audio clips, such as a character, a location or a voice. You can fix the last frame so the clip ends exactly where you want, or continue a video you already have.

It is Pika's tuned version of MiniMax's H3 video model, the same base as the orbit add-on in issue 117. It finishes a clip in six steps instead of the usual dozens, which keeps it fast. The free version renders at 480p or 768p. Pika's paid service goes up to 4K.

02 How to try it

Open the Pika H3 Flex Space and use its builder. Add a shot, describe what happens and how the camera moves, then add a second shot with a cut time. Type one line of dialogue and a music cue, keep 480p and 5 seconds for the first run, and generate. Watch whether the lips match the line and whether the sound changes at the cut. Then upload a photo of a product or a person as the opening frame and run the same script. The Space runs on Hugging Face's shared free GPUs, so a free account gets a limited number of takes. Running it yourself means a data center card and over 200 GB of downloads.

03 Caveat

In clips with several cuts, faces and outfits can drift from shot to shot, and lip sync can slip on longer lines, so start with two or three shots. The model also inherits MiniMax's H3 community license, which does not cover using the model or its videos in the United States, the European Union, the United Kingdom or South Korea.

04 What you can do with it

  • Turn one product photo into a ten second spot with camera moves, a voiceover and music.
  • Storyboard a scene with sound before you film anything.
  • Bring a drawn character to life and give them a line to say.
  • Continue a clip you already have in the same look.

Try the demo

View model page

Read this issue on neuralexpedition.com →