FAQ about creating with AI

Pick a video model, describe the scene in plain words and send it. The studio shows the exact price before you send, and the finished clip waits in your library. You can start from words alone or from a photo you upload, and many models add sound too. Right now an 8 second clip at 720p in Veo 3.1 Lite costs ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound, and the complete guide to AI video walks through every step

Upload the photo as the start image in a video model that supports it, and describe in words what moves and how the camera travels. The model starts from your photo and animates it, and some models also take an end image for the clip to finish on. Almost all our video models take a start image, for example Veo 3.1 Lite, Kling 3.0 and Seedance 1.5 Pro. Every detail and tip is in the image to video guide

Choose a model with built-in sound, and it creates the dialogue, effects and ambience together with the picture. Veo 3.1, Kling 3.0, Seedance 2.5 and Wan 3.0 work this way, and each model’s note says whether it has sound. To make a character say a specific line, put the line in quotes in the prompt. No video model here officially supports Hebrew speech, so a Hebrew voiceover is made separately with a speech model

One generation gives you anywhere from a few seconds to half a minute, depending on the model. Veo 3.1 makes up to 8 seconds, Kling 3.0 up to 15, LTX-2.5 Fast up to 20, and Seedance 2.5 and Wan 3.0 up to 30 seconds. A longer video is built from several clips joined in an editor, and the guide to longer AI video shows how

Choose the 9:16 aspect ratio, the vertical phone format. Most of our video models offer it, including Veo 3.1 Fast, Kling 3.0 and Seedance 2.0. If you start from a photo, a vertical one works best, since some models fit the clip’s shape to your image. How to choose a video model covers which models suit Reels and TikTok

In text to video the model builds the whole scene from your prompt, and in image to video it starts from a photo you upload and animates it. A photo gives you full control of the first frame, such as a real product or a character you already designed, while text suits a scene you want the model to invent. Most of our video models do both, and some also take reference images that guide a character or style without being the first frame. The full explanation is in the image to video guide

The start image is the clip’s first frame and the end image is the frame it finishes on, and the model invents the motion in between. That lets you fix exactly where a scene begins and ends, such as a closed box that opens. In most models an end image works only together with a start image, and it is offered in Veo 3.1, Kling 3.0 and Seedance 1.5 Pro, among others. In the studio the end image field appears only on models that take one

Yes, in some models. Kling 3.0 and LTX-2.3 Pro offer 4K. In MiniMax H3, Gemini Omni Flash 1.1, Veo 3.1 Lite and Seedance 2.0 4K is an upscale of a lower resolution result, and in Veo 3.1 and Veo 3.1 Fast with sound on and no reference images, 1080p and 4K are made here by enlarging a 720p clip. 4K costs a lot more: an 8 second clip in Veo 3.1 costs ₪5.8658.58 crVeo 3.1, 720p · 8s · with sound at 720p and ₪8.9189.05 crVeo 3.1, 4K · 8s · with sound at 4K

Upload the clip to Wan 2.7 Video Edit and write what to change, such as a new style, different clothes or an item added to a character. You can attach a reference photo, of the clothing for example, and choose to keep the original sound. The clip should be 2 to 10 seconds long, and an upload from the studio can be up to 4 MB. Kling O3 takes a video as a reference, but it makes a new shot from it rather than editing your clip

Yes. You can create with Seedance 2.0 here from Israel, with a Hebrew interface and a price in shekels before every generation. Right now a 5 second clip at 720p costs ₪4.2242.18 crSeedance 2.0, 720p · 5s · with sound. There are also Seedance 2.0 Fast and Seedance 2.0 Mini for cheaper drafts, and Seedance 2.5 for clips of up to 30 seconds

Give the model reference images of the character, and it keeps the look the same in each new clip. The number varies by model: up to 3 images in Veo 3.1, up to 7 in Kling O3, up to 9 in Seedance 2.0 and up to 30 in Seedance 2.5. Another way is to start every clip from the same start image of the character. The guide to a consistent character has the full method

Pick an image model, describe what you want to see and send it, and the image is saved to your library the moment it is ready. The exact price shows before you send, and each extra image in the same request is charged on its own. Right now an image in Nano Banana 2 costs ₪0.191.88 crNano Banana 2, 1K · 1 image, and a medium quality image in GPT Image 2 costs ₪0.222.21 crGPT Image 2, 1K · 1 image · Medium quality. For a prompt that works, see how to write an image prompt, and the full guide, with choosing a model and Hebrew, is how to create an image with AI

Upload the image to an editing model and write what to change, such as "swap the background for a beach" or "put a red coat on the character". Nano Banana Pro, GPT Image 2, FLUX.1 Kontext [pro] and Seedream 5.0 Pro all edit from a written instruction. Ask for one change at a time and say what should stay as it is. The maker of FLUX Kontext notes that many edits in a row can add visual flaws, and when that happens, go back to the original image. More on editing in words in the image guide

Upload all the images to a model that takes several inputs, and write what to take from each one and how to combine them, for example "the character from the first image sits in the room from the second". Nano Banana Pro and Nano Banana 2 take up to 14 images, GPT Image 2 up to 16 and FLUX.2 [pro] up to 8. Each model’s maximum shows next to the images field in the studio

Attach a clear photo of the face as a reference image in every generation, and say in the prompt that it is the same person. Nano Banana Pro is built to keep the look of up to 5 people, and Seedream 4.5 keeps the facial features, light and color tone of your reference. Google notes that character consistency is not always perfect even in Nano Banana Pro, so check every image. The guide to a consistent character has the full method

Choose a model that supports a transparent background, such as GPT Image 2.5 Flare or GPT Image 2.5 Sunburst, and set the background to transparent. The file has to be PNG or WEBP, since JPEG does not keep transparency. In GPT Image 2 transparency is still in preview at its maker, so the result is not guaranteed. It is handy for a logo or a product you place on another design later

Choose a model that offers 4K and set the resolution before you send. Nano Banana Pro, Nano Banana 2 and Seedream 4.5 all go up to 4K. In Nano Banana 2 a 1K image costs ₪0.191.88 crNano Banana 2, 1K · 1 image right now, and a 4K one ₪0.424.22 crNano Banana 2, 4K · 1 image. Google notes that small text is often blurry at 1K in Nano Banana Pro, so an image with a lot of text is worth a higher resolution

For Hebrew text, in our experience the new GPT Image models are the best choice, such as GPT Image 2 and GPT Image 2.5 Flare, and Nano Banana Pro also writes good Hebrew. For a poster with English text there are more options, such as Ideogram 4.0, built for posters and graphic design, or FLUX.2 [flex], which its maker describes as specialized in text inside images. For a logo on a transparent background, GPT Image 2.5 Flare fits. Either way, check every letter before you use the image

Paste the text into a speech model that officially supports Hebrew, pick a voice and you get an audio file. Eleven v3, Eleven v4, Eleven v4 Turbo, MiniMax Speech 2.8 HD and MiniMax Speech 2.8 Turbo all list Hebrew among their official languages. In MiniMax Speech set Language boost to Hebrew, and in Eleven v4 you can set the language code to he. The full comparison is in the Hebrew voiceover guide

Yes, but not in every model. Eleven v3, Eleven v4 and Eleven v4 Turbo have Hebrew on ElevenLabs’ official language list, while Eleven Multilingual v2, Eleven Turbo v2.5 and Eleven Flash v2.5 do not

Five of our models officially support Hebrew, and we have not yet run a structured listening test between them, so there is no proven single winner. Eleven v3 and Eleven v4 suit highly emotive narration, Eleven v4 Turbo and MiniMax Speech 2.8 Turbo cost less than their full versions, and MiniMax Speech 2.8 HD lets you pick an emotion and control speed and pitch. The best way is to make the same short line in two or three models and choose by ear. Right now 1,000 characters cost ₪0.474.69 crEleven v3, 1,000 chars in Eleven v3 and ₪0.474.69 crMiniMax Speech 2.8 HD, 1,000 chars in MiniMax Speech 2.8 HD

In Eleven v3 you write tags in English inside the text, such as [whispers] for a whisper or [laughs] for a laugh, and they change how the line is read. Eleven v4 takes tags like these too, and even background sounds such as [light rain]. In MiniMax Speech 2.8 HD you pick an emotion for the whole read, such as happy or sad, and you can add tags like (laughs) or (sighs) inside the text. ElevenLabs notes that Eleven v3 takes more trial and error with wording than its earlier models

Eleven v4 is the newer model, from September 2026, with 90+ languages against 70+ in Eleven v3, and both include Hebrew. ElevenLabs built v4 to read tone, pacing, emotion and character from the text, and in both you can steer the read with tags inside the text. With us both take up to 5,000 characters per request, and right now 1,000 characters cost ₪0.474.69 crEleven v3, 1,000 chars in v3 and ₪0.383.75 crEleven v4, 1,000 chars in v4. There is also Eleven v4 Turbo, a faster and cheaper version of v4

Speech models charge by the number of characters, and ElevenLabs puts 5,000 characters at about 5 minutes of speech, roughly 1,000 a minute. Right now 1,000 characters cost ₪0.191.88 crEleven v4 Turbo, 1,000 chars in Eleven v4 Turbo, ₪0.474.69 crEleven v3, 1,000 chars in Eleven v3 and ₪0.474.69 crMiniMax Speech 2.8 HD, 1,000 chars in MiniMax Speech 2.8 HD, all of which support Hebrew. The exact price for your text shows before you generate, based on its length

Pick a music model, describe the genre, tempo and mood, and in models that take lyrics, add those too. Eleven Music v2.5, Lyria 3 Pro, Lyria 3.5 and MiniMax Music 3 make a full song with vocals, and Lyria 3 makes a 30 second track. In MiniMax Music 3 you write the lyrics yourself, while the Lyria models write them from your description. Right now a one minute song in Eleven Music v2.5 costs ₪2.8128.12 crEleven Music v2.5, 60s

You can try, but none of our music models officially supports Hebrew vocals. Lyria 3 and Lyria 3 Pro officially sing in 8 languages without Hebrew, and Lyria 3.5, Eleven Music v2.5 and MiniMax Music 3 publish no language list. Google says Lyria 3.5 writes lyrics in the language of your prompt, but Hebrew is not confirmed. If Hebrew lyrics matter, make a short track first and listen before you invest in a full song

Lyria 3 makes 30 second tracks, Lyria 3 Pro makes full songs of up to about 3 minutes with verses, chorus and bridge, and Lyria 3.5 is Google’s newest, with more natural vocals and better lyrics. In Lyria 3 Pro you can write what happens at each moment with timestamps, and in Lyria 3.5 the song’s length and tempo come from your prompt. Right now Lyria 3 costs ₪0.191.88 crLyria 3, One track, Lyria 3 Pro costs ₪0.383.75 crLyria 3 Pro, One track and Lyria 3.5 costs ₪0.474.69 crLyria 3.5, One track

In Eleven Music v2.5 and MiniMax Music 2.6 turn on the Instrumental only switch, and in Eleven Music it guarantees a track with no vocals. In the Lyria models, write in the prompt that you want an instrumental with no vocals. Name the instruments, tempo (BPM) and mood, and in Eleven Music also set the length in seconds to match your video. Right now a 30 second track in Lyria 3 costs ₪0.191.88 crLyria 3, One track

Split the lyrics into sections with tags such as [Verse] and [Chorus], and put each line on its own row. In MiniMax Music 3 and MiniMax Music 2.6 the lyrics go in their own field, and the prompt carries the genre, tempo, instruments and mood. In Lyria 3 Pro you can write what happens at each moment with timestamps, such as [0:00 - 0:10] for a quiet intro. Naming a tempo (BPM) and a key also helps, since Lyria and MiniMax Music 2.6 read them from the prompt

Describe in plain sentences what you see: who or what is in the center, where it happens, the light, the style and the camera angle. One concrete detail beats a lot of general words like "beautiful" or "amazing". When the image needs text, write the exact words in quotes and say where they go. If the result is off, change one thing at a time and keep the wording that worked

Describe the scene like directions for a camera operator: what we see, what moves, how the camera travels, the light and the style. In models with sound, add what we hear, such as street noise or soft music, and put dialogue in quotes. Google’s official prompting guide for Veo 3.1 suggests a five part formula, and stresses quotes for dialogue and cues for effects and ambience. The video prompt guide walks through it with examples

Write who speaks, and put the exact line in quotes, for example: the woman looks at the camera and says "Welcome to our store". You need a model with built-in speech, such as Veo 3.1, Kling 3.0 with sound on, or Kling 2.6. In a clip of a few seconds, keep the line short enough to be said in full. No video model here officially supports Hebrew speech, and Kling translates dialogue in a language outside its list into English

You can ask in words for moves such as a slow push in, a pull back, a sideways track, an orbit around the subject, following a character or a camera that stays still. MiniMax documents bracketed commands for Hailuo 2.3, such as [Pan left] or [Push in]. In LTX-2.5 Pro you pick a ready made camera move from a list, such as Move in, Move out or Rise, and Seedance 1.5 Pro has a fixed camera switch. Ask for one clear move per shot

Because every generation starts from a random point, so even the very same prompt gives a slightly different result each time. To get results closer to each other, write a more detailed prompt and use the same start image or the same reference images in every generation. When a result works, keep its prompt and change only one detail at a time. If you create through the API, MCP or CLI, you can also set a seed number in models that support it, and the same number with the same prompt and settings gives a very similar result

For most video and image models English is the safer choice, since the makers that publish a language list leave Hebrew out of it. Google confirms full support for English in Veo 3.1 and says other languages have not been evaluated, and Wan and Qwen Image officially support Chinese and English. Text that should appear inside an image goes in Hebrew, in quotes, and speech models such as Eleven v3 get the Hebrew text itself. More in Hebrew in AI models

Start from what the clip needs: speech and sound, length, a consistent character or a low price. For a scene with dialogue there is Veo 3.1, for clips of up to 30 seconds Seedance 2.5 or Wan 3.0, for a consistent character a model that takes reference images, and for many cheap tries Veo 3.1 Lite. Every model has a short note on what it does best, and the exact price shows before each generation. The full comparison is in how to choose a video model

No single model always wins, but for social media pick one with 9:16, built-in sound and a low price per clip. Veo 3.1 Lite has all three, and an 8 second clip in it costs ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound right now. Seedance 1.5 Pro and Kling 2.6 also offer 9:16 and sound. In most cases 720p is enough for social feeds, so there is no need to pay for 4K

It depends on the task, and no single model leads at everything. For Hebrew text inside an image, in our experience the new GPT Image models are the best choice, for editing and blending many images Nano Banana Pro, for fast low cost images Nano Banana 2 Lite, and for text heavy infographics Seedream 5.0 Pro. On an independent voting leaderboard from October 2026, GPT Image 2.5 Flare ranks 2nd of 82 for text to image. The two front runners are compared in Nano Banana Pro vs GPT Image 2

We have not run our own comparison yet, but makers and independent rankings give clear signs. MiniMax says Hailuo 2.3 handles complex body movement and subtle facial expressions, and Seedance 2.0 keeps a character’s face and clothing consistent through the clip. On an independent voting leaderboard from September 2026, MiniMax H3 ranks 1st of 48 for image to video. The surest way is to make the same short scene in two or three models and compare

Yes, and it is the quickest way to find the model that fits your style. Every model here runs on the same balance, and when you switch models the exact price updates at once. To save, compare in the budget versions at a short length, for example in Veo 3.1 Lite or Seedance 2.0 Mini, and move to the full model only for the final version. Use the exact same prompt so the comparison is fair

These are names makers give to versions of the same model: Fast, Turbo and Lite are usually quicker or cheaper versions, and Pro is usually the quality version. For example, Google calls Veo 3.1 Lite the most cost-effective in the Veo 3.1 family, and 1,000 characters cost ₪0.191.88 crEleven v4 Turbo, 1,000 chars in Eleven v4 Turbo against ₪0.383.75 crEleven v4, 1,000 chars in Eleven v4. The name does not tell the whole story: Wan 3.0 Prime is a faster version that costs more than Wan 3.0, not less. So compare the live price and the note of each version before you choose

According to OpenAI, the Sora app closed on April 26, 2026 and its API on September 24, 2026, so it can no longer be used. The closest alternatives here are Veo 3.1 for cinematic scenes with dialogue, Kling 3.0 for clips of up to 15 seconds, Seedance 2.5 for clips of up to 30 seconds and Gemini Omni Flash 1.1 for fast clips with a soundtrack. They all run on the same balance, with no subscription. The details are in the guide to Sora’s shutdown

Nano Banana Pro is Google’s quality model for complex jobs such as infographics and blending many images, and Nano Banana 2 is the fast one, which Google describes as high quality at a mainstream price. Both take up to 14 images and go up to 4K, and Nano Banana 2 also offers long formats such as 4:1 and 8:1 for banners. In our experience both write good Hebrew inside images. Right now a 1K image costs ₪0.191.88 crNano Banana 2, 1K · 1 image in Nano Banana 2 and ₪0.424.22 crNano Banana Pro, 1K · 1 image in Nano Banana Pro

Veo 3.1 is the better pick for cinematic scenes with dialogue and built-in sound, and Kling 3.0 when you need a longer clip, up to 15 seconds against 8. In Veo sound is on by default and you can give up to 3 reference images, while in Kling sound is optional and you can pick any length from 3 to 15 seconds. Neither officially supports Hebrew speech. Right now 8 seconds at 720p in Veo 3.1 cost ₪5.8658.58 crVeo 3.1, 720p · 8s · with sound and 10 seconds at 1080p in Kling 3.0 cost ₪4.2242.18 crKling 3.0, 1080p · 10s · no sound, and Veo 3.1 vs Kling 3.0 has the full comparison

Seedance 2.0 is strong at fast, complex motion such as sports and action, and takes up to 9 reference images plus a sample clip and a sample audio file. Kling 3.0 suits lip synced dialogue and keeps the subject stable while the camera moves, but here it takes no reference images. Seedance 2.0 has sound on by default, while in Kling 3.0 it stays off until you turn it on. Right now 5 seconds at 720p cost ₪4.2242.18 crSeedance 2.0, 720p · 5s · with sound in Seedance 2.0 and ₪1.6416.41 crKling 3.0, 720p · 5s · no sound in Kling 3.0

It depends on what the clip needs: Veo 3.1 for dialogue and cinematic sound in clips of up to 8 seconds, Seedance 2.0 for complex motion and a character kept from up to 9 reference images, and Kling 3.0 for lip synced dialogue in clips of up to 15 seconds. None of the three officially supports Hebrew speech. For up to 30 seconds in one generation there is also Seedance 2.5. The detailed comparison is in how to choose a video model

Seedance 2.5 makes clips of up to 30 seconds in one generation, against 15 in Seedance 2.0, and takes up to 30 reference images against 9. ByteDance says 2.5 keeps the looks and voices of several characters consistent through the clip. In both, sound is generated with the picture, and both also take a sample clip and a sample audio file. Right now 5 seconds at 720p cost ₪4.2242.18 crSeedance 2.0, 720p · 5s · with sound in 2.0 and ₪5.4254.18 crSeedance 2.5, 720p · 5s · with sound in 2.5, and Seedance 2.0 vs 2.5 has the full comparison

All three make 4 to 8 second clips with sound, and the main differences are price, dialogue and rankings. Veo 3.1 is the full model, Veo 3.1 Fast costs less, and Google calls Veo 3.1 Lite the most cost-effective in the family. Google documents up to 3 reference images and 4K for Veo 3.1 and Fast and not for Lite, but here you can pick both in Lite too. Right now an 8 second clip at 720p costs ₪5.8658.58 crVeo 3.1, 720p · 8s · with sound, ₪1.4114.06 crVeo 3.1 Fast, 720p · 8s · with sound and ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound respectively. The breakdown is in Veo 3.1 vs Fast vs Lite

GPT Image 2 is the better pick for text inside images, Hebrew included, and for text heavy visuals such as infographics and posters. Nano Banana Pro is built for blending many images and keeping the look of up to 5 people. In our experience both write good Hebrew, and the new GPT Image models are especially strong at it. Right now an image costs ₪0.222.21 crGPT Image 2, 1K · 1 image · Medium quality in GPT Image 2 at medium quality and ₪0.424.22 crNano Banana Pro, 1K · 1 image in Nano Banana Pro at 1K, and Nano Banana Pro vs GPT Image 2 has the full comparison

Wan 3.0 is the newest, from August 2026 and still marked preview by Alibaba, and makes clips of up to 30 seconds with speech, effects and background music. Wan 2.7 makes up to 15 seconds, can continue an existing clip and can time the motion to an audio file you upload. Wan 2.6 makes 5, 10 or 15 seconds and can split a clip into several shots. To edit an existing clip in words there is Wan 2.7 Video Edit, and all of them officially take prompts in Chinese and English

Kling 3.0 is the main model, with 3 to 15 second clips, up to 4K and sound you can switch on. Kling O3 is built to keep characters and style from up to 7 reference images, or to make a new shot based on a reference video. Kling 2.6 is the previous generation, with 5 or 10 second clips that have sound and speech on by default. There is also Kling 3.0 Turbo, at 720p or 1080p, from text or a single start image

GPT Image 2.5 Flare is the small, fast version of GPT Image 2.5, and OpenAI says its quality is comparable to GPT Image 2. GPT Image 2.5 Sunburst is the base model, which OpenAI says gives higher image quality and suits work where editing precision matters most. Both offer transparent backgrounds and five quality levels, from low to max. On independent voting leaderboards from October 2026, Flare ranks 2nd for text to image and Sunburst 1st for image editing

FLUX.2 [pro] is the main model for creating and editing, including blending up to 8 images. FLUX.2 [flex] is, per its maker, specialized in text inside images and small details, and lets you tune the step count and how closely it follows the prompt. FLUX.2 [dev] is an open weight model that here makes images from text only, with advanced settings and up to 4 images per request. Hebrew text inside images has not been tested in the FLUX versions here, so for Hebrew text a model with the "Good Hebrew" mark is the safer pick. Right now a 1K image costs ₪0.060.57 crFLUX.2 [dev], 1K · 1 image in Dev, ₪0.121.18 crFLUX.2 [pro], 1K · 1 image in Pro and ₪0.242.35 crFLUX.2 [flex], 1K · 1 image in Flex

Yes. The whole interface works in Hebrew from right to left, with English one tap away, and every model has a short plain note on what it does best

The models marked "Good Hebrew". The mark appears when the model’s official language list includes Hebrew, when our own tests passed, or, for some image models, from our own experience using them. For a voice model it means good reading of Hebrew text, and for an image model correct Hebrew text inside the image. The models page can show just those

There is no official confirmation. For Veo 3.1, the version you can use here, Google confirms full support for English and says other languages have not been evaluated and may work with varying results. We have not tested Hebrew speech either. For a Hebrew voiceover it is safer to make the clip on its own and add a voice from a speech model that supports Hebrew, such as Eleven v3

Not officially. In Kling 3.0 the official speech languages are Chinese, English, Japanese, Korean and Spanish, and dialogue in any other language is translated into English. In Kling 2.6 speech is officially Chinese and English, and other languages are translated into English there too. For a Hebrew voiceover, make the voice separately in a speech model that supports Hebrew and join them in an editor

Make the clip and the voiceover separately and join them in an editor. The clip can come from any video model, with or without sound, and the voiceover from a speech model that officially supports Hebrew, such as Eleven v3 or MiniMax Speech 2.8 HD. You can also upload the voiceover to Wan 2.7 as a driving audio file, and it times the lip movement to it, but there is no official confirmation for Hebrew and we have not tested it. It all runs on one balance, and the Hebrew voiceover guide has the details

Some do, but most models have no official confirmation for Hebrew. The speech models Eleven v3, Eleven v4 and MiniMax Speech 2.8 HD officially support Hebrew text, since reading it is their job. Video and image models are different: Google fully confirms only English in Veo 3.1, and Wan and Qwen Image officially support only Chinese and English. So write the scene description in English, and put text that should appear in the image in Hebrew, in quotes

Upload a good photo of the product as the start image and describe the motion and the camera, for example "the camera slowly circles the bottle on a marble table". Seedance 1.5 Pro suits product clips with voiceover and camera moves, and it has a fixed camera switch. Kling notes that when you animate a photo, the result depends heavily on the source photo’s resolution, so use a sharp one. Right now 5 seconds in Seedance 1.5 Pro cost ₪0.828.21 crSeedance 1.5 Pro, 720p · 5s · with sound, and the image to video guide has more ideas

Upload a photo of the product to an editing model and ask for a new background, light and setting, while saying that the product itself must stay as it is. Nano Banana Pro, GPT Image 2 and Seedream 5.0 Pro take several input images, so you can attach the product from more than one angle. Make one image first, check that the product came out true to the original, and only then make a series. Right now an image in Nano Banana Pro costs ₪0.424.22 crNano Banana Pro, 1K · 1 image. The full method is in the image guide

Make a few short clips, each with one job such as the product, the use or the message, and join them in an editor with text and music. The price follows the model and the length of each clip: right now 8 seconds in Veo 3.1 Fast cost ₪1.4114.06 crVeo 3.1 Fast, 720p · 8s · with sound, and 10 seconds in Kling 3.0 cost ₪4.2242.18 crKling 3.0, 1080p · 10s · no sound. A Hebrew voiceover is made separately, and 1,000 characters in Eleven v3 cost ₪0.474.69 crEleven v3, 1,000 chars. The editing itself and buying the ad space happen outside the site

Prepare one fixed set of reference images, of the product or the character, and attach it to every generation in the campaign. For images Nano Banana 2 takes up to 14 reference images, and for video Seedance 2.5 takes up to 30. Keep one fixed style description and paste it into every request, so every result looks like part of the same series. The guide to a consistent character covers the whole method

Before every generation you see the exact price in credits, and that is the amount that leaves your balance, not an estimate. Changing the length, resolution or image count updates it at once

The model you choose and its settings: clip length, resolution, sound, number of images, quality and the length of the text read aloud. Every model page under models and prices has a price table for its settings, and the guide to prices explains each of them

Each generation costs the price shown before it. The price list itself follows the daily dollar rate and changes in model prices. If the price goes up just before you send, nothing is sent until you confirm the new price

It depends on the model, the length and the resolution, and the exact price shows before every generation. Right now an 8 second clip at 720p costs ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound in Veo 3.1 Lite, ₪1.4114.06 crVeo 3.1 Fast, 720p · 8s · with sound in Veo 3.1 Fast and ₪5.8658.58 crVeo 3.1, 720p · 8s · with sound in Veo 3.1. A 10 second clip at 1080p in Kling 3.0 costs ₪4.2242.18 crKling 3.0, 1080p · 10s · no sound. Every model page and the price guide have price tables by length and resolution

Right now an 8 second clip at 720p costs ₪5.8658.58 crVeo 3.1, 720p · 8s · with sound in Veo 3.1, ₪1.4114.06 crVeo 3.1 Fast, 720p · 8s · with sound in Veo 3.1 Fast and ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound in Veo 3.1 Lite, sound included. A higher resolution costs more: an 8 second clip in Veo 3.1 costs ₪5.9859.76 crVeo 3.1, 1080p · 8s · with sound at 1080p and ₪8.9189.05 crVeo 3.1, 4K · 8s · with sound at 4K. Every model page has a table showing how length and resolution change the price

In Kling 3.0 the price follows the seconds and the resolution: right now 5 seconds at 1080p cost ₪2.1121.09 crKling 3.0, 1080p · 5s · no sound, and 10 seconds cost ₪4.2242.18 crKling 3.0, 1080p · 10s · no sound. At 720p the same 5 seconds cost ₪1.6416.41 crKling 3.0, 720p · 5s · no sound, and at 4K ₪7.8578.50 crKling 3.0, 4K · 5s · no sound. Turning on sound adds to the price: 5 seconds at 1080p with sound cost ₪3.1631.64 crKling 3.0, 1080p · 5s · with sound. The full price table is on the model page

It depends on the model, the quality and the resolution, and each image in a request is charged on its own. Right now an image costs ₪0.090.94 crNano Banana 2 Lite, 1 image in Nano Banana 2 Lite, ₪0.191.88 crNano Banana 2, 1K · 1 image in Nano Banana 2 at 1K, and in GPT Image 2 ₪0.060.57 crGPT Image 2, 1K · 1 image · Low quality at low quality and ₪0.606.00 crGPT Image 2, 1K · 1 image · High quality at high quality. The exact price shows before every generation, and the price guide explains what changes it

No model here makes a full minute in one generation, so a minute is built from several clips and the price is their sum. For example, a 30 second clip at 720p in Wan 3.0 costs ₪7.0370.30 crWan 3.0, 720p · 30s · no sound right now, so a minute from two clips costs twice that. In Veo 3.1 Lite a minute takes about 8 clips of 8 seconds, each costing ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound. In practice, leave room for retries since not every clip comes out right the first time, and the guide to longer AI video shows how to join them

It changes with length, resolution and the dollar rate, so the studio’s model list is sorted from cheapest to most expensive by the price of one generation at base settings. Examples from right now: an 8 second clip at 720p in Veo 3.1 Lite costs ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound, and 5 seconds at 720p in Seedance 1.5 Pro cost ₪0.828.21 crSeedance 1.5 Pro, 720p · 5s · with sound. Google calls Veo 3.1 Lite its most cost-effective video model. When you compare, work it out per second, since lengths differ between models

Not here. Every generation is charged to your balance, and the exact price shows before you send, so you know in advance what it will cost. To try it for a small amount, start with a low cost model: right now an 8 second clip at 720p in Veo 3.1 Lite costs ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound. There is no monthly plan, credits do not expire, and a failed generation gives the credits back. When paying per generation beats a plan is explained in subscription or pay per generation

Veo 3.1 Lite is cheaper. Right now an 8 second clip at 720p costs ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound in it, against ₪1.4114.06 crVeo 3.1 Fast, 720p · 8s · with sound in Veo 3.1 Fast. Google calls Lite the most cost-effective model in the Veo 3.1 family and says it runs at the same speed as Fast. Here you can pick reference images and 4K in both, and Fast is the better pick when the character speaks, since Google describes dialogue for it

Yes. Every price on the site shows in credits and in shekels, and the shekel amount includes VAT. One credit is worth ₪0.10, so 10 credits make one shekel. The price list moves with the daily dollar rate, but the price shown before a generation is the price of that generation

A credit is how you pay on the site. One credit is worth ₪0.10 including VAT, so 10 credits make one shekel. Every price on the site shows in credits, with the sum in shekels beside it

No. You buy credits in advance and pay only for what you create. There is no plan to pick, no recurring charge and no subscription to cancel

No. Unused credits stay in your balance and do not expire while your account exists. Nothing resets at the end of the month

For anyone who creates as needed, paying per generation usually works out better, since on many plans unused credits are lost when the month resets. On a plan, the real price of a generation is the plan price divided by what you actually used: use half the credits and every generation costs twice as much. A plan can pay off for someone who uses a large plan to the last credit every month. The full math with examples is in subscription or pay per generation

Usually because of costly settings that are easy to miss: high resolution, long clips, many images per request or maximum quality. Resolution alone can raise the price a lot: a 5 second clip in Kling 3.0 costs ₪1.6416.41 crKling 3.0, 720p · 5s · no sound at 720p and ₪7.8578.50 crKling 3.0, 4K · 5s · no sound at 4K. Here the exact price shows before every generation, a failed one gives its credits back, and unused credits do not expire. To save, draft in a budget version at a low resolution, as the price guide explains

All its credits go back to your balance, automatically and with nothing to ask for. A generation that has not finished within two hours is checked one last time: if the result is ready it is saved to your library, and if not, it is closed and its credits come back. Every refund shows in your activity

You do not need to do anything: a generation that has not finished within two hours is checked one last time. If the result is ready it is saved to your library, and if not, the generation is closed and all its credits go back to your balance. Meanwhile you can leave the page and keep creating, and the status of every generation shows on the jobs page. The details are in a failed or stuck generation

Every result is saved in your personal library the moment it is ready. The first 1 GB is free, and each extra GB costs ₪10 a month, in a small daily charge taken from your balance

New generations and uploads are blocked until you top up or delete files. The files themselves are not deleted for this reason, and you can still see them

Yes. Every file has a direct link on our site, and anyone with the link can open the file, so send it only to people who need it. Deleting the file cancels the link

For as long as your account exists. Every result is saved in your personal library and stays there until you delete it or the account, and files are not deleted even when your balance cannot cover storage. The files are kept in cloud storage in the European Union, and the first 1 GB is free. The details are in the privacy policy

Here your results do not depend on the model staying available: every result is copied to our storage and saved in your library the moment it is ready. So if a model stops working, the files you already made with it stay in your library, and you can carry on in another model from the same balance. When Sora closed, OpenAI said its data is deleted after an export window, so keep your own copy of any file that matters too

Yes. Every account includes API keys, a command line tool (CLI) and an MCP connection at no extra cost, with the same models, the same prices and the same library as the website. Through the API and MCP every generation carries a price cap, max_credits, and one that costs more is not sent and nothing is charged

No. Everything works in the studio on the website. The API, CLI and MCP are extras for anyone who wants to create from software, the terminal or an AI assistant such as Claude

Create an API key on the developers page, and use it to get a price estimate, send a generation and check its status. The estimate is free, and every generation request must carry a price cap, max_credits, so a generation that costs more is never sent and nothing is charged. Every model and option in the studio is available through the API, with the same prices and the same library. Input files, such as a start image, go in as https links

Yes. With our API key you can create in Veo 3.1, Veo 3.1 Fast and Veo 3.1 Lite without opening a cloud account of your own. You pay from the same credit balance as on the site, and every request carries a price cap. Right now an 8 second clip at 720p in Veo 3.1 Lite costs ₪0.707.03 crVeo 3.1 Lite, 720p · 8s · with sound

MCP (Model Context Protocol) is an open standard that lets an AI assistant such as Claude use outside tools from inside the chat. Through our MCP connection the assistant can pick a model, get a price estimate, create images, video and voice and check your balance. The connection works with your API key, and every generation through it carries a price cap. Setup for Claude Code and Cursor is on the developers page

Connect Claude to our MCP connection with an API key, then simply ask in the chat, for example "make an 8 second clip in Veo 3.1 Lite". The connection tells Claude to get a price estimate first, send the generation with a price cap and return a link to the finished file. In Claude Code the setup is one terminal command, shown on the developers page. The full guide is creating images and videos from Claude

Deletion is final: the files, keys, balance and history are deleted, and credits left in the balance are lost. Payment records are kept after the deletion, as set out in the privacy policy. To get a refund for unused credits, ask for it before deleting