From one jar photo to a lit hero shot
Three steps, and one tell to look for.
- 01
Upload the real candle
A clean photo of the jar, label, and wax. Unlit is fine, the models add the flame and the light.
- 02
Describe the light, not just the scene
Say how the room should feel, for example warm evening light on a wood table with the jar glowing from within.
- 03
Check for the melt pool
Compare results and look for a molten pool around the wick and a halo on the surface. That is what separates a lit candle from a candle on a shelf.
What you can restage
Keep the jar and label, change the room and the hour.
- Time of day and warmthBright morning shelf styling, or a dim evening scene where the flame is the main light source.
- Surface and propsWood, linen, stone, or wool, with botanicals and spices that signal the scent without naming it.
- Lit or unlitAn unlit jar for a clean catalogue packshot, or a lit one with a melt pool for the in-use hero.
- Season and occasionAutumn and gifting scenes for Q4, fresh and airy styling for spring launches.
Which model for which candle shot
Verdicts from running one candle brief through all four. Every model drew the flame; only some rendered its light.
OpenAI
GPT Image 2
The most convincing internal glow. It lit the whole frosted jar warm from the inside, which is the shot most candle brands actually want.
ByteDance
Seedream 4.5
The premium macro hero, and the cheapest of the photoreal picks. The only one that rendered a genuine melted wax pool around the wick, with warm light cast onto wood and wool.
Nano Banana 2
The coziest full lifestyle scene, integrating the flame's glow into a styled setting with props like lavender and cinnamon.
Black Forest Labs
FLUX.2 Pro
The prettiest still jar, but watch the flame: no visible wick, no melt pool, and almost no cast light. Save it for shots where you will light the scene yourself.
Shipping these to a storefront
Renders are high-resolution and licensed for commercial use, so they can go straight onto product pages, marketplaces, and seasonal campaigns. One candle-specific note: if your jar carries printed scent names, burn times, or safety copy, treat that as regulated text and verify it at full size rather than trusting a generated label, and keep an honest photo of the unlit product on the listing.
The same candle, four models, one prompt
First-hand renders from our candle test. Look at the wick, the pool, and the light on the surface rather than at the flame itself.
AI candle photography questions
What candle makers ask before trusting a catalogue to this.
Can AI render a realistic flame?
Every model in our test drew a convincing flame shape, because flames are common in training data. That is precisely why the flame is not where you choose a model. What separates a real candle image from a fake one is whether the flame behaves like light: a glow through the frosted glass, a warm halo on the surface, and a melted pool where the heat has been working. GPT Image 2 and Seedream 4.5 rendered that physics. FLUX.2 Pro drew the flame and forgot the light, and the eye catches that instantly even when it cannot name it.
Is the glass jar a problem?
Not anymore, and this is worth flagging because candle-specific tools have historically warned that reflective glass is hard, since the vessel blends poorly into a swapped background. That is a background-replacement limitation, not a generation one. In from-scratch generation all four models handled frosted and clear glass fine. The vessel is not where these models struggle. The lighting is.
What makes a candle shot look genuinely lit?
The melt pool. A freshly poured candle has a flat, even wax surface; a lit one has a small molten pool around the wick. Seedream 4.5 rendered that pool, which is a large part of why its image reads as a candle in use rather than a candle on a shelf. If you want a lit and cozy mood, look for the pool first and the flame second.
Will it keep my exact product?
Yes, and this is the part that matters most. You upload a clean photo of your real product, and the models restyle the scene around it, so the label text, shape, color, and finish stay yours instead of being reinvented. Different models hold fidelity differently on the hard cases (frosted glass, small printed labels, reflective metal), which is exactly why comparing several on your actual product beats trusting one pipeline. Upload the clearest, best-lit shot you have for the most faithful result.
Which model should I use for candles?
Match it to the mood. GPT Image 2 for a frosted jar that glows from within. Seedream 4.5 for a premium macro with a real melted pool, at a fraction of the cost. Nano Banana 2 for the coziest full lifestyle scene. FLUX.2 Pro makes the prettiest still jar but the most decorative flame, so save it for when you will light the scene yourself.
Can I use several models on one product?
There is no single best model, and that is the point. Different models win at different briefs, so Masonry gives you all of them in one workspace and lets you compare results side by side. For candles and home fragrance, Nano Banana Pro is the best all-rounder for photoreal results with faithful layout, Seedream is fast and detailed for exploring options, FLUX gives the most control for a consistent style, and GPT Image is great for precise edits. You pick the winner for each shot instead of being locked into one engine.
Related guides
Go deeper on the models and tools behind these shots.
Shoot your candle line without a studio
Upload one jar, compare the models on it, and keep the shot whose flame actually lights the room.
