Google Flow Image to Video: Frames, Ingredients and Credits

Google Flow turns a still image into a clip in two ways. Frames to Video uses your image as the literal first frame, or uses two images as the first and last frames, and animates what happens between them. Ingredients to Video treats images as references, so a character, product or setting stays the same from one clip to the next. As of November 2026, both video models in Flow accept images: Omni Flash 1.2 and Veo 3.1 Lite. A 10-second Omni clip at 720p costs 30 credits by our arithmetic, and a Veo 3.1 Lite clip costs 10. Start with Omni on the web. Use Veo 3.1 Lite as the cheap baseline, knowing its Ingredients mode only runs at 8 seconds.

Frame or ingredient: decide what the image is for

The choice comes down to one question. Is your image the shot itself, or a thing that should appear in the shot? Flow Help describes frames as the way to animate an image, build a transition, or get a video that starts and ends on specific pictures. Ingredients do a different job: they keep the same character and key objects consistent between clips while the model composes each new shot on its own. A product photo you want to see slowly rotate is a frame. The same product sitting on a kitchen counter in three different scenes is an ingredient.

Both modes share the same prompting rules. Flow Help asks for a prompt that complements the visual inputs rather than contradicting them, and it recommends naming the frames or ingredients inside the prompt. For products and people used as references, it advises a plain or segmented background, which keeps the model from copying clutter into every clip. Flow also recommends English prompts; other languages work, with quality that varies. Our Google Flow prompt guide covers camera wording and dialogue in more depth. Here the focus stays on the image side.

Frames to Video: one image, or a first and a last

With Omni Flash 1.2, Frames to Video works in both aspect ratios, from 3 to 10 seconds without a plan and up to 15 seconds on Google AI Plus, Pro or Ultra. Omni's Frames to Video is web only: Flow Help says it is not in the mobile app yet ^[1]. Veo 3.1 Lite handles a first frame alone, or a first and last frame together, at 4, 6 or 8 seconds in 720p. That makes Veo 3.1 Lite the option for a quick animated still when you do not need more than 8 seconds.

The first-and-last setup is where frames earn their place. DeepMind's prompt guide for Gemini Omni suggests defining the opening frame, then the camera movement or transition, then what the final frame reveals. It names orbits, zoom transitions and loops as good fits. In practice that means a before-and-after reveal, a push-in from a wide room shot to a close-up, or a product turning to face the camera. Spend the prompt on motion. Name the frames, then describe the move. The two images already tell the model where to begin and where to land.

Longer sequences are built clip by clip. Flow lets you save any frame of a video to your project as an image, then use it as an ingredient, a start frame or an end frame in a later generation. The last frame of one clip can become the first frame of the next, and Scenebuilder then arranges, reorders and trims the clips into one sequence. Google's own model card for Veo 3.1 Lite, read in November 2026, reports it won 47.2% of image-to-video comparisons against Veo 3.1 Fast, a model Flow retires on December 5, 2026. That is just under half. It is Google's own test.

Ingredients, characters and @me

Ingredients to Video runs on Omni Flash 1.2 from 3 to 10 seconds for everyone and up to 15 seconds on paid plans. On Veo 3.1 Lite it runs at 8 seconds only, so a 4- or 6-second Veo clip has to come from text or frames. If you are on a paid plan and need more length, extension adds 3 to 10 seconds to an Omni clip, guided by up to 3 image references. It is web only. Veo Extend was retired on November 5, 2026; extending now needs a paid plan and Omni Flash 1.2.

Characters turn loose ingredients into something reusable. A character bundles visual and audio references, so the face, clothing and voice stay consistent ^[2]. You upload or generate 1 or 2 images, and a character must contain at least one image to work. In the prompt you call it with @ and its name. Your own avatar is called with @me. Voice references take a single speaker, are tagged @Voice, and can only be added to generations that use ingredients on Omni Flash ^[3].

Custom voices start from a preset voice that you restyle by describing the change. Flow Help notes that some Omni features and voice options are limited to certain regions. Avatars carry more conditions. They are unavailable in the EEA, the UK and Switzerland, creation is finished on a phone through a QR code, and avatar data cannot be shared through Flow Tools or public project links. For a brand mascot or a recurring presenter, a character built from two clean images on a plain background is the sturdier base.

Making the frames with Nano Banana

You do not need a camera to feed Frames to Video. Flow's image models in November 2026 are Nano Banana Pro, Nano Banana 2.1 and Nano Banana 2 Lite, and generating images uses no Flow credits. Each model has a daily cap instead. Nano Banana Pro allows up to 20 images a day without a plan, 100 on Google AI Plus or Pro, and 500 on Ultra. Nano Banana 2.1 and 2 Lite share one pool of 400 a day without a plan and 1,000 on paid plans. Nano Banana 2 Lite is the default at no charge.

Each cap resets 24 hours after your first generation with that model, and failed or blocked images do not count against it. That shapes a sensible order of work. Generate the first frame, the last frame and the character sheets as images, iterate on them until they look right, and spend credits only on the video. DeepMind's Omni prompt guide says references created with Nano Banana, or taken from real life, help keep characters, objects and environments consistent. Image-to-image editing is listed among Flow's modes, so a frame can be adjusted before it is animated.

Which model takes which input, and what it costs

Credits are charged per generation, not per request, and one request can produce several generations ^[4]. By our arithmetic, a 10-second 720p Omni clip costs 30 credits, a 15-second 720p clip costs 60 and a 15-second 360p clip costs 20. The free tier's 50 daily credits cover five Veo 3.1 Lite clips, five 10-second Omni clips at 360p, or about 16 seconds of 720p Omni video. Google AI Pro's 1,000 monthly credits cover 33 ten-second 720p Omni clips or 100 Veo 3.1 Lite clips.

Omni at 360p costs one third of the credits of 720p, which makes it the right setting for testing whether two frames blend well. On Google AI Plus, Pro or Ultra, upscaling a 1-to-10-second Omni clip to 720p or 1080p costs 0 credits. Test cheap, then upscale. Credits for failed or safety-blocked generations come back within one hour. A clip that works but misses the brief is still charged. Daily and monthly credits do not roll over; our Google Flow credits guide covers the rules. Google retires Veo 3.1 Fast, Veo 3.1 Quality and Omni Flash 1.1 in Flow on December 5, 2026; until then they stay available at standard credit rates (Veo vs Omni).

Where image-to-video stops

The ceiling is short. One generation runs 15 seconds at most on Omni with a paid plan, 10 without one, and 8 on Veo 3.1 Lite. Anything longer is extension plus Scenebuilder, one clip at a time. The cheapest plan that unlocks 15-second clips and extension is Google AI Plus at $4.99 a month, with 200 Flow credits a month. Google AI Pro costs $19.99 a month or $199.99 a year and adds 1,000 Flow credits a month. Free users may find video generation unavailable roughly 2 PM to 5 PM UTC, while images keep working.

Some photos will not animate at all. In every region, Flow does not support generating prominent people, editing uploaded videos of identifiable minors, or editing what people say in uploaded videos. Flow is for users 18 and over, age-verified, in supported countries, and a VPN does not change that. Uploaded-video editing is not available in the EEA, the UK or Switzerland either. Plan for the web. Omni's Frames to Video and extension both live there, and Google recommends a desktop Chromium browser for Flow.

The bottom line

Flow's image-to-video comes down to two choices: whether your image is a frame or an ingredient, and whether Omni Flash 1.2 or Veo 3.1 Lite fits the clip. Omni covers every mode, from 3 to 10 seconds for everyone and up to 15 on paid plans, at 1 or 3 credits a second (2 or 6 a second for seconds 11 to 15); its Frames to Video and extension are web only. Veo 3.1 Lite costs a flat 10 credits a clip, but its ingredients only run at 8 seconds. Build frames and characters with Nano Banana, which spends no credits, draft at 360p, and keep each clip within the 15-second ceiling.