If your AI tool has no negative prompt box, you exclude things by describing what should be there instead: “an empty street at dawn” rather than “no cars”. FLUX.2 works this way: Black Forest Labs says “FLUX.2 does not support negative prompts.”1 So do ChatGPT Images, Runway, Google Flow and Seedance 2.5, and each vendor says what to do in its own guide. OpenAI and ByteDance also allow a short “no …” or “Avoid …” line; Runway warns that negative wording can backfire.
- No box is normal now. Many newer image and video tools take one prompt only.
- Name what is there instead. “A bald man”, not “a man with no hair”. “Sharp focus throughout”, not “no blur”.
- Say what must stay the same. When editing, list what to keep, and repeat it every time.
- A short “no” line is fine in ChatGPT and Seedance, because their makers use one in their own examples.
- Never on Runway. Runway says negative phrasing may produce the opposite result.
Tell a friend “don’t think of a red car” and a red car is the first thing they picture. Image models can stumble the same way: the word you wanted gone is still a word in the prompt. So instead of naming what you do not want, describe the scene where it simply is not there. Not “no people”, but “an empty beach”.
Which tools have no negative prompt field?
The plain point: most of the big chat and video tools take one prompt, with no separate “leave this out” box.
| Tool | Negative field? | What the vendor says to do |
|---|---|---|
| ChatGPT Images | No separate box. | Write exclusions into the prompt, and list what must stay the same. |
| FLUX.2 | No. | Describe what you want, not what you don’t want. |
| Runway Gen-4 and later | No, and negatives may backfire. | Positive phrasing only. |
| Google Flow | None in the prompt settings. | Describe the scene in detail; describe exclusions as part of the scene. |
| Seedance 2.5 | No. | A short “Avoid” clause at the end of the prompt. |
| Gemini image models (Nano Banana) | None documented. | Describe what you want, not what you don’t. |
| Kling | Only on the legacy API. | Negatives as sentences inside the prompt. |
For tools that do have a box, such as Stable Diffusion, SDXL and Midjourney, see negative prompts for Stable Diffusion and SDXL and the model-by-model negative prompts for every AI model. Stability AI’s API also has a negative_prompt field,2 and Veo on Google Cloud takes a negativePrompt.3 Google’s Imagen is different: on a Vertex AI page Google now marks as no longer updated, it calls Imagen’s negative prompt “a legacy feature”, missing from Imagen models from imagen-3.0-generate-002 on.4 And Kling keeps its field only on the legacy API, where it still recommends sentences in the prompt.5
Three ways to exclude something without a negative field
The plain point: describe the replacement, lock what must stay, and only then, where the vendor allows it, add one short exclusion line.
- Name what is there instead. Google’s advice for its Gemini image models: instead of “no cars”, write “an empty, deserted street with no signs of traffic.”6
- Lock what must stay. When you edit an image, say what to change and what to keep, and repeat the keep list on every follow-up. OpenAI’s image guide says to “repeat the preserve list on each iteration to reduce drift.”7
- Add one short exclusion line, only where it is welcome. ChatGPT Images and Seedance use one in their makers’ own examples. Runway does not: leave it out there.
ChatGPT Images
The plain point: describe the image, then add a short line of exclusions and a list of what must not change.
OpenAI’s help page says to describe the image you want, and its developer guide lists no negative-prompt setting. It tells you to state exclusions in the prompt itself.89
To edit one part, OpenAI’s help page says to name that area in your prompt.8 More: the ChatGPT Images cheat sheet.
FLUX.2
The plain point: there is no negative prompt at all. Rewrite every “no” as a description.
Black Forest Labs, which makes FLUX, says “FLUX.2 does not support negative prompts” and gives two rewrites: say “sharp focus throughout” instead of “no blur”, and describe an “empty scene” instead of “no people”.10
More: the FLUX cheat sheet.
Runway
The plain point: positive words only. On Runway a “no” can pull in the thing you banned.
Runway’s image guide gives the classic rewrite: not “a man with no hair” but “a bald man”.11 Its video guide does the same for motion: not “No camera movement” but “Locked camera.”12
More: the Runway cheat sheet.
Google Flow
The plain point: Flow’s prompt settings have no negative box. Describe the scene, and describe the missing thing as part of it.
Google’s Flow help page lists the prompt box settings (aspect ratio, number of outputs, model and generation length), and none of them is a negative prompt. Its advice is to describe what you want.13 Google’s prompting guide for Veo, the video model inside Flow, says to describe what you want to leave out, such as “a desolate landscape with no buildings or roads”.14
Notice the “no” is still there, but inside a description of the scene, as in Google’s own example. More: the Veo cheat sheet and the Nano Banana cheat sheet.
Seedance 2.5
The plain point: no field, but ByteDance’s own template keeps a short “Avoid” clause at the very end.
The template puts what must survive first and what must not happen second.15
More: the Seedance cheat sheet.
Instead of this, write this
The plain point: every common “no” has a positive twin. Copy the right-hand column.
| Instead of | Write | Why it works |
|---|---|---|
| no blur | sharp focus throughout | Black Forest Labs’ own rewrite (above). |
| no people | an empty scene; an empty beach at dawn | Black Forest Labs’ own rewrite (above). |
| a man with no hair | a bald man | Runway’s own example (above). |
| no camera movement | locked camera, the camera remains still | Runway’s own example (above). |
| no cars | an empty, deserted street | Google’s own example (above). |
| no text on the bottle | a plain unbranded bottle with a blank label | This site’s own: name the clean surface. |
| no plastic skin | visible pores, slightly uneven skin tone, unretouched | This site’s own: name the texture. |
| no extra people | she stands alone in the frame | This site’s own: say how many. |
| no cartoon look | documentary photograph on 35mm film | This site’s own: name the medium. |
| no flicker | steady, even lighting through the whole shot | This site’s own: name the steady state. |
The same rewrites, sorted by problem (skin, hands, text, duplicates, flicker and more): negative prompts by problem.
Questions people ask
Where is the negative prompt field in Google Flow?
There isn’t one in the settings Google documents for the Flow prompt box: aspect ratio, number of outputs, model and generation length. Describe the scene you want, and word any exclusion as part of it, as Google’s Veo guide does (“a desolate landscape with no buildings or roads”; sources above).
Why does writing “no people” sometimes add people?
Because the model can treat the named thing as something to include. Google’s own example: in “a rainy city street at night with no people”, the model may read “people” as something to include.4
Does ChatGPT understand “no” in an image prompt?
OpenAI’s own guides use short lines such as “No extra text, no watermarks, no unrelated logos,” so a brief exclusion line is the vendor’s advice there (source above). Still check every result, and describe the scene you want first.
No negative box? Describe the scene where the unwanted thing is simply absent, list what must stay the same, and add a short “no” or “Avoid” line only in tools whose makers use one (ChatGPT Images, Seedance). On Runway, keep every word positive.
Related on this site: what a negative prompt means · prompt meaning · text-to-image meaning · image-to-video meaning · what the vendors actually say.
Related on this site: Free Negative Prompt Builder for AI Images and Video
- 1Black Forest Labs, FLUX.2 prompting guide, read at source 3 Nov 2026“No negative prompts: FLUX.2 does not support negative prompts.” ↑
- 2Stability AI, API reference, read at source 13 Oct 2026negative_prompt, “A blurb of text describing what you do not wish to see in the output image. This is an advanced feature.” ↑
- 3Google Cloud, Generate videos from text prompts, read at source 13 Oct 2026negativePrompt, “A string value that describes content that you want to prevent the model from generating.” ↑
- 4Google Cloud, Omit content using a negative prompt (Vertex AI documentation; the page carries the banner “Vertex AI documentation is no longer being updated”), read at source 13 Oct 2026negative prompts are “a legacy feature” and “are not included with the Imagen models starting with imagen-3.0-generate-002 and newer”. ↑afor the prompt “a rainy city street at night with no people”, the page notes the model may read “people” as something to include rather than omit. ↑b
- 5Kling AI, Text to Video API (legacy), read in a browser 12 Oct 2026“It is recommended to supplement negative prompt via negative sentences within positive prompts”. ↑
- 6Google Cloud, Gemini image generation best practices, last updated 9 Oct 2026, read at source 12 Oct 2026“Instead of saying ‘no cars’, describe the scene positively by saying, ‘an empty, deserted street with no signs of traffic.’” ↑
- 7OpenAI, GPT Image Generation Models Prompting Guide, read at source 12 Oct 2026“repeat the preserve list on each iteration to reduce drift.” ↑
- 8OpenAI Help Center, Images in ChatGPT, read in a browser 12 Oct 2026“To create an image, describe the image you want ChatGPT to create.” ↑a“If you want the edit to apply to a specific area, include that area in your prompt.” ↑b
- 9OpenAI, Image prompting, read at source 12 Oct 2026“State exclusions such as unwanted text, logos, or watermarks.” ↑
- 10Black Forest Labs, FLUX.2 prompting guide, read at source 12 Oct 2026“FLUX.2 does not support negative prompts.” “Focus on describing what you want, not what you don’t want.” “Instead of ‘no blur,’ say ‘sharp focus throughout.’” “Instead of ‘no people,’ describe an ‘empty scene.’” ↑
- 11Runway, Gen-4 Image Prompting Guide, read in a browser 12 Oct 2026“Negative prompts, or prompts that describe what shouldn’t appear in the image, are not supported in Gen-4 Images. Including a negative prompt may result in the opposite happening. ❌ a man with no hair ✅ a bald man” ↑
- 12Runway, Gen-4 Video Prompting Guide, read in a browser 12 Oct 2026“Negative phrasing is not supported and may produce unpredictable or even opposite results. ❌ No camera movement. The camera doesn’t move. NO MOVEMENT ✅ Locked camera. The camera remains still.” ↑
- 13Google, Create videos in Google Flow, read at source 12 Oct 2026“Describe the scene you want to create in detail. Be specific about the subject, action, environment, lighting, and style.” “Your text prompt should complement, not contradict, your visual inputs.” ↑
- 14Google Cloud blog, Ultimate prompting guide for Veo 3.1, 16 Oct 2025, read at source 12 Oct 2026“To refine your output, describe what you wish to exclude. For example, specify ‘a desolate landscape with no buildings or roads’ instead of ‘no man-made structures’.” ↑
- 15Dreamina (ByteDance), Seedance 2.5 prompt guide, 18 Aug 2026, read at source 12 Oct 2026“Preserve [identity, wardrobe, logo, object geometry].” “Avoid [unwanted cuts, camera behavior, artifacts, extra objects, text].” “Keep them short, concrete, and grouped at the end. Do not write a long generic blacklist that competes with the scene description.” ↑
Read at source 12 Oct 2026 unless marked: OpenAI, Image prompting and GPT Image Generation Models Prompting Guide · OpenAI Help Center, Images in ChatGPT (read in a browser) · Black Forest Labs, FLUX.2 prompting guide · Runway, Gen-4 Image Prompting Guide and Gen-4 Video Prompting Guide (read in a browser) · Google, Create videos in Google Flow · Google Cloud blog, Ultimate prompting guide for Veo 3.1, 16 Oct 2025 · Google Cloud, Gemini image generation best practices, last updated 9 Oct 2026, and Omit content using a negative prompt (re-read 13 Oct 2026; marked “Vertex AI documentation is no longer being updated”) · Dreamina (ByteDance), Seedance 2.5 prompt guide, 18 Aug 2026 · Kling AI, Text to Video API (legacy) (read in a browser).
The copy-paste prompts, the “red car” comparison and the rows marked “this site’s own” are this site’s own ways of explaining, not claims from these sources. Tool features change: check the vendor page before relying on a setting.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK