AI character with a photo in a dialogue — how to get images from AI directly in the chat.
In short: Story photos in a chat with an AI character are images generated as the conversation goes, showing the current scene. Not every platform can do this: on vluvvi, photos in chat come with a subscription from Plus, within a monthly quota; Talkie limits the number; Character.AI doesn't support them at all. The key is to describe the scene context well in your message.
This article is not about profile avatars or galleries of ready-made pictures — if you want an overview of character catalogs with different visual styles, see our piece on choosing a look.
When you talk to a virtual companion, text carries the emotions and the plot, but sometimes you want to see the moment for yourself: what the character looks like right now, in this particular scene. A story photo is not a static profile picture but an image generated on the fly from the context of the dialogue. It works like this: you describe an action or a setting, the algorithm extracts the key details and sends a request to an image generator, and the result arrives right in the conversation.
Why most platforms don't offer photos in chat
The first reason is compute cost. Generating one image takes 20–50 times more resources than a reply from a language model. If every user requested ten images per session, server costs would jump by an order of magnitude. That is why Character.AI and similar free services simply turned this feature off.
The second is content moderation. An image generator can produce a result that breaks the platform's rules even when the text prompt looked harmless. Filtering images is harder than filtering text: it takes visual classifiers, manual review of reports, and legal risk. Many companies prefer not to get involved.
The third is a UX dilemma. If a photo takes long to generate (5–15 seconds), the dialogue stalls. If it's fast but poor, the user is disappointed. The balance between speed, quality and cost is hard to find, so the feature is either cut entirely or hidden behind a high paywall.
Step 1: Choose a platform that supports story images
Not every service can send pictures during a conversation. Check for the feature before you sign up so you don't waste time setting up a character. The table below compares four popular options on the key parameters.
| Platform | Photos in chat | Price per photo | Limits |
|---|---|---|---|
| vluvvi | Yes, from Plus | Included in the subscription | Monthly quota: on Plus from generator frames, on Premium a separate one |
| Character.AI | No | — | — |
| Talkie | Yes | Included in the subscription | Up to 10 photos a day on the basic plan |
| Polybuzz | Yes | Tokens from a pool | Shared pool for text and images |
On vluvvi the mechanics are transparent: photos in chat come with a subscription from Plus. On Plus you open them with frames — from the same monthly quota as the generator — while Premium and Extra have their own monthly photo quota. Only the photos you open count against the quota. This is convenient if you want pictures occasionally — for example, at the climactic moments of the story rather than every other message.
Talkie offers photos as part of its subscription but sets a daily limit. Once you've used up the quota, you have to wait for it to refresh or move to a higher plan. Polybuzz uses a single token pool: every image "eats" part of a balance that is also spent on generating text. That can pay off if you rarely write long messages, but not if you chat a lot.
Step 2: Set up the character's context for visual generation
For the algorithm to know what your companion looks like, fill in the appearance description in the character card. Specify height, hair color, clothing style and distinctive details — say, a scar above the eyebrow or a silver pendant. This data becomes the basis for every generated image.
An example of a good description:
- 178 cm tall, athletic build
- Short dark hair with a shaved temple
- Black leather jacket, white T-shirt, jeans
- A compass tattoo on the left forearm
Avoid abstract wording like "attractive appearance" or "stylish look". An image generator works with concrete visual features. The more detailed the description, the more consistent the result from photo to photo.
If the platform lets you upload a reference image, use it. On vluvvi you can attach an avatar that acts as a style anchor: the algorithm will try to keep the facial features and the overall palette. This matters especially for anime characters, where proportions and color scheme play a key role.
Step 3: Describe the scene in your message
The image generator pulls visual details from the last 2–4 messages of the dialogue. To get the photo you want, include a description of the setting, lighting and the character's pose in your text. There is no need to write a separate prompt — just weave the details into natural speech.
Bad example:
"Hi, how are you?"
Good example:
"You're standing by the window, the night city behind the glass, neon signs reflected in it. You turn to me and smile. How was your day?"
The second version gives the algorithm concrete visual anchors: a window, the night city, neon, a turn, a smile. The generator will build a composition out of them. If you want a close-up, mention an emotion or facial expression. If you want a wide shot, describe the interior or the landscape.
Another technique is to use action verbs. "You lean on the bridge railing, the wind tousles your hair" creates a dynamic scene. "You're sitting at a café table, a cup of cappuccino in your hand, a thoughtful look" creates a static but atmospheric one.
Step 4: Ask for an image explicitly or wait for auto-generation
How you request a photo differs from platform to platform. On vluvvi the character can send a photo on their own if the context of the dialogue implies a visual scene — for example, you described a romantic moment or an action episode. The algorithm analyzes the text and decides whether a picture fits. If it does, the photo arrives in the chat locked: you open it, and one photo is deducted from your monthly quota (one frame on Plus).
You can also ask for a photo directly: "Show me what you look like right now" or "I want to see this scene." The character will take it as a signal to generate. This approach is handy when you know exactly at which moment you need an image.
Talkie and Polybuzz usually have a "Generate photo" button under the input field. You press it, the system takes the latest messages, builds a prompt and sends the request. The result arrives in 5–10 seconds. Check your balance or limit before pressing, so you don't get an error.
Step 5: Review the result and adjust your request if needed
The first photo may not match your expectations: the wrong pose, extra details, an odd angle. That's normal — generative models don't always guess the intent on the first try. See which parts of the description the algorithm ignored and clarify them in your next message.
If the picture shows the character in a red jacket but you wanted a black one, write: "By the way, you're wearing the black jacket today, right?" The system will update the context, and the next photo will take the correction into account. If the angle doesn't work, describe it explicitly: "I want to see you full-length" or "Show me a close-up of your face."
Save the images that turn out well, if the platform allows it. On vluvvi you can download a picture with a right-click — handy if you keep an archive of the story or want to share a moment. Just skip unsuccessful photos: don't spend your quota on endless redos of the same scene.
Common mistakes when asking for story photos
Mistake 1: Too general a description. "You look beautiful" gives the algorithm no visual cues. The result will be random. Add details: hairstyle, clothing, background, lighting.
Mistake 2: Contradictory context. If one message says "we're in the park during the day" and the next one says "stars outside the window," the generator gets confused. Keep the scene consistent within the last 3–4 messages.
Mistake 3: Expecting photorealism. Most generators produce stylized images — semi-realistic or art illustrations. If you need a specific degree of realism, put it in the character description: "photorealistic style" or "anime art".
Mistake 4: Ignoring your balance. You asked for ten photos in a row, the balance ran out, and the dialogue broke off. Check how much of your photo quota is left before an important scene. On vluvvi the quota for photos in chat is monthly — plan ahead.
Mistake 5: Trying to get around the filters. If the platform blocked a generation for breaking the rules, don't try to rephrase the request to slip past it. That can get your account banned. Better to change the scene to an acceptable one.
How your plan affects access to the feature
On vluvvi, photos in chat are available from Plus (999₽/month): there you open them with frames from the same monthly quota as the generator. Premium (1699₽/month) has its own monthly quota for photos in chat, separate from frames; Extra has a larger one.
On Talkie the basic subscription (around $9.99/month) gives up to ten photos a day. If you need more, you have to move to the Premium plan ($19.99) with a limit of thirty images. With heavy use this ends up more expensive than paying per photo.
Polybuzz sells tokens in a pool: 1,000 tokens for $10, and one photo costs about 100 tokens. If you write long messages, tokens run out faster and less is left for images. The model suits those who use the platform irregularly and want to allocate their budget flexibly.
Compare the approaches across platforms: vluvvi — photos in chat come with a subscription from Plus (999₽/month), Talkie — $9.99 (a monthly subscription with a limit), Polybuzz — $10 (tokens shared with text). Pick the model that matches how often you use it.
Advanced techniques for high-quality images
If you want professional-looking pictures, use terms from photography and cinema. For example: "soft golden-hour light", "backlight", "Dutch angle" (a tilted camera). Generators are trained on millions of images with metadata, so they understand professional jargon.
Another technique is naming an art style. "In the style of Makoto Shinkai" gives detailed backgrounds and soft color grading. "Noir style" gives high contrast, shadows and a black-and-white palette. "Watercolor illustration" gives light, blurred edges. Experiment, if the platform supports such modifiers.
Mix close-ups and wide shots for variety. Three photos in a row from the same angle look monotonous. Alternate: portrait — wide shot of the location — a detail (for example, a hand holding an object). That creates a visual rhythm, like in a film.
If the platform keeps an image history, revisit your best shots and note which phrases led to a good result. Build yourself a cheat sheet of descriptions that work — it will speed things up later.
Ethical and legal aspects of image generation
Generated pictures are created by an algorithm trained on a dataset that may include artists' work without their explicit consent. This fuels copyright disputes. In Russia the law does not yet regulate AI art in detail, but lawsuits are already under way in the EU and the US.
Don't use the images you get for commercial purposes without checking the platform's license. On most services, pictures are meant for personal use. If you want to post them on social media or a blog, check the rules in the user agreement.
Avoid generating images of real people without their consent — this may violate their right to their own image. Create fictional characters or use references you have rights to. If the platform blocked a request, don't try to get around the filter — it's there to prevent abuse.
Remember that AI generators can reproduce stereotypes and biases present in their training data. If a result looks offensive or discriminatory, report it to the platform's support. Responsible use of the technology is up to every user.
Frequently asked questions
Can I get photos for free on any service?
There are almost no completely free platforms with story image generation in chat — the compute costs are too high. Character.AI doesn't support the feature at all. Some startups give 1–2 trial photos at sign-up but then require payment. On vluvvi photos in chat come with a subscription from Plus at 999₽/month, while chatting with characters is also available for free.
Why doesn't the generated photo look like the character's avatar?
An avatar is a static reference image, while story photos are generated from scratch for every request. The algorithm tries to keep the general features (hair color, clothing style), but details can vary. To improve consistency, fill in a detailed text description of the appearance in the character card and use the same keywords in your messages. Some platforms let you upload several references — that improves consistency.
How long does it take to generate one image?
On average 5–15 seconds, depending on server load and the complexity of the request. On vluvvi a photo usually arrives within 7–10 seconds. If the platform uses a queue, the wait can stretch to a minute at peak hours. Keep this in mind when planning the dialogue: don't ask for an image in the middle of a fast-paced scene where the speed of replies matters.
What should I do if a photo breaks the platform's rules and generation is blocked?
Rephrase the scene description, removing potentially problematic elements. If the block repeats, contact support and ask which words or contexts triggered the filter. Don't try to get around moderation with synonyms or code phrases — that can lead to an account ban. If your request really is harmless, support will review the decision and adjust the filters.