Models

GPT Image 2.6 Guide: Flare vs Sunburst, GIFs, SVGs and 12 Real Tests

ChatCut Team
32 min read
GPT Image 2.6 Guide: Flare vs Sunburst, GIFs, SVGs and 12 Real Tests
Contents
  1. How Do GPT Image 2.6 Flare, Sunburst, Quality, and Complexity Compare?
  2. How Do Looping GIFs Work in GPT Image 2.6? (Tests 3 & 4)
  3. Can GPT Image 2.6 Export Editable SVGs and Transparent PNG Stickers? (Tests 5 & 6)
  4. How Accurate Is Multi-Line and Bilingual Chinese Typography? (Tests 7 & 8)
  5. How Steady Are Reference Edits, Product Swaps, Characters, and Sketch Renders? (Tests 9–12)
  6. All 12 Tests Compared: Credits, Timing, and When to Switch Models
  7. Frequently Asked Questions

GPT Image 2.6 (released by OpenAI as ChatGPT Images 2.6 on November 4, 2026) adds native looping GIFs, editable 1K SVG vectors, a three-level complexity planning control, and steadier reference-locked photo edits across two identically priced variants: Flare for speed and Sunburst for strict spatial and typographic precision. Instead of paraphrasing launch notes, we ran 24 real generations across 12 commercial, editorial, architectural, and motion tests in ChatCut, timing every job and verifying every settled credit charge (6.30 credits total). Browse the finished visuals and live looping GIFs below first, then jump to the settings, copyable prompts, and fixes behind each test.

Glossy high-fashion magazine cover titled SOLSTICE Issue 14 featuring an editorial portrait with clean coverlines reading THE NEW SILHOUETTE and AUTUMN / WINTER.
Editorial cover layout from ChatCut's GPT Image 2.6 showcase (4:5): multi-tier masthead and coverlines rendered cleanly over photography.
High-speed commercial beverage ad of a frosted silver-and-citron YUZU CLUB sparkling citrus can exploding out of a crystal-clear water crown splash and crushed ice beside a halved yuzu under Zero Sugar All Bite copy.
Test 1 · GPT Image 2.6 Sunburst (4:5, 1K high, 54s median, 0.22 credits): exact three-tier ad copy plus halved yuzu anchored inside the base splash.
Looping 16:9 animated GIF of a dark midnight recording studio booth window where a red ON AIR neon sign flares bright vermilion while four equalizer bars rise from blue to amber-red and settle back.
Test 3 · Live looping GIF on GPT Image 2.6 Flare (16:9, 9 frames at 180 ms, 1280×720 px, 0.49 credits): locked studio booth with cyclic neon and meter pulse.
Golden-hour medium-format architectural photograph of a cantilevered concrete and frameless-glass cliffside villa with a turquoise infinity pool, two teak loungers, cypress trees, and twin island peaks at sunset.
Test 12 · Sketch-to-photo on GPT Image 2.6 Sunburst (16:9, 1K high, high complexity, 0.21 credits): rendered directly from a black-and-white ink perspective drawing.
Looping 1:1 transparent-background animated GIF of a brushed-silver and signal-orange cassette tape die-cut sticker labeled SIDE A with spinning white spools and two cobalt musical notes popping out.
Test 4 · Transparent-alpha sticker GIF on GPT Image 2.6 Flare (1:1, 8 frames at 150 ms, 960×960 px BGRA, 0.40 credits): drops straight over video tracks.
Four unbranded filmmaker die-cut vinyl stickers (a 35mm rangefinder camera, cobalt studio headphones, a SCENE 01 clapperboard, and a glowing red REC tally light) composited over a dark checkerboard to prove RGBA transparency.
Test 6 · Transparent PNG sticker sheet on GPT Image 2.6 Flare (1:1, 1056×1056 px RGBA, 0.27 credits), shown over dark checkerboard to verify inner cutouts.
Morning window-blind shadows across honed travertine stone with a frosted celadon glass perfume bottle beside a cotton letterpress card printed with KYOTO RAIN 京都朝露, HINOKI & WHITE TEA 扁柏与白茶, and AUTUMN EDITION 秋季限定 50 ML.
Test 8 · Bilingual English + Simplified Chinese campaign card on GPT Image 2.6 Sunburst (4:5, 1K xhigh at 0.38 credits; 0.22 credits at high).
Fluted crystal perfume bottle labeled NOIR VÉTIVER standing on rain-wet black volcanic basalt hit by a warm diagonal spotlight casting golden glass caustics beside vetiver roots and eucalyptus leaves.
Test 10 · Studio-to-location product swap on GPT Image 2.6 Sunburst (4:5, 1K high, low complexity, 0.26 credits): 100% label and fluted-glass preservation.

How Do GPT Image 2.6 Flare, Sunburst, Quality, and Complexity Compare?

YUZU CLUB sparkling citrus commercial ad generated with GPT Image 2.6 Flare at low quality, showing readable three-tier text and softer water splash droplets.
Flare · low quality (22s, 681 output tokens, 0.11 credits): half the cost of high, all three text strings intact, softer splash edges.
YUZU CLUB commercial beverage ad generated with GPT Image 2.6 Flare at high quality, with crisp strobe droplets and a halved yuzu floating above the top rim of the can.
Flare · high quality (21s median, 1,454 output tokens, 0.22 credits): crisp strobe rim light, halved yuzu floated above the rim on all 3 runs.
YUZU CLUB commercial beverage ad generated with GPT Image 2.6 Sunburst at high quality, placing the halved yuzu directly inside the base ice splash as prompted.
Sunburst · high quality (54s median, 1,453 output tokens, 0.22 credits): placed the halved yuzu inside the base splash on all 3 runs.

OpenAI launched ChatGPT Images 2.6 and its System Card on November 4, 2026, building on April’s GPT Image 2 and September’s 2.5 split. In ChatGPT’s consumer chat interface, everyday prompts route through Flare while harder multi-constraint layouts escalate to Sunburst automatically. In ChatCut, you pick GPT Image 2.6 Flare or GPT Image 2.6 Sunburst directly and control resolution, aspect ratio, quality, complexity, and output format up front.

To measure real speed and cost without network batch distortion, we ran the exact same 4:5 commercial beverage strobe prompt seven times one after another: three alternating runs on Flare high, three on Sunburst high, and one draft check on Flare low.

Variant & setting (4:5, 1K)RunsGeneration timeOutput image tokensSettled creditsWhat differed on screen
GPT Image 2.6 Flare · high quality321s median (20s, 21s, 21s)1,4540.223/3 spelled all copy right; floated halved yuzu above the can for dramatic symmetry
GPT Image 2.6 Sunburst · high quality354s median (53s, 54s, 58s)1,4530.223/3 spelled all copy right; obeyed “around the base” literally + sharper pull-tab metal
GPT Image 2.6 Flare · low quality122s6810.11Half the output tokens of high (vs ~1/9th in 2.5); clean ad copy, slightly softer ice

Three practical rules came out of Test 1:

  1. Flare and Sunburst cost the exact same credits at matching settings. On 4:5 at 1K high, both settled at 0.22 credits; on 16:9 at 1K high, both settle at 0.15 credits. Start on Flare when iterating lighting or framing in 15 to 22 seconds, and switch to Sunburst when exact spatial prepositions (“beside the base”, “third shelf from left”) matter.
  2. Low quality is now a usable half-price layout draft. In 2.5, low quality dropped to roughly one-ninth of high’s token budget and mangled small lettering. In 2.6, low uses nearly half of high’s output tokens (681 vs 1,454 on 4:5) and kept “ZERO SUGAR. ALL BITE.” and “SPARKLING CITRUS · 330 ML” readable at 0.11 credits.
  3. Aspect ratio changes token count. At 1K high without references, 16:9 settled at 0.15 credits (1,002 tokens), 3:4 at 0.20 credits (1,344 tokens), 4:5 at 0.22 credits (1,453 tokens), and 1:1 at 0.27 credits (1,796 tokens).
Try this promptOpen in ChatCut
High-speed commercial beverage advertising photography on medium format, 4:5 vertical. A matte frosted silver-and-citron slim aluminum can standing upright on wet black obsidian stone, hit by razor-sharp studio dual-strobe rim lighting. Dynamic crystal-clear sparkling water crown splash and crushed translucent ice cubes exploding around the base alongside a fresh halved Japanese yuzu fruit with glistening pulp. Bold clean condensed Swiss grotesque typography printed crisply on the can: 'YUZU CLUB' at center, 'SPARKLING CITRUS · 330 ML' below. Clean bold white studio campaign headline at the top of the frame reading 'ZERO SUGAR. ALL BITE.' Deep charcoal-black studio background, hyper-crisp water droplets, zero motion blur, commercial campaign print finish.
Result

Test 1 benchmark prompt · 0.22 credits at 1K high (Flare 21s median, Sunburst 54s median) · 0.11 credits on Flare low.

What Complexity Low vs High Changes on Multi-Callout Layouts (Test 2)

Overhead industrial knolling photograph of six CNC aluminum field-recorder parts on grey concrete generated with Sunburst at low complexity, showing an asymmetric 1-4-1 layout and 90-degree bent callout lines.
Test 2 · Sunburst · complexity: low (~34s, 1,002 output tokens, 0.15 credits): accurate labels, asymmetric column grouping with bent leader lines.
Overhead industrial knolling photograph of six CNC aluminum field-recorder parts on grey concrete generated with Sunburst at high complexity, arranged in a balanced 3x2 grid with straight uncrossed callout lines.
Test 2 · Sunburst · complexity: high (~52s, 1,002 output tokens, 0.15 credits): balanced 3×2 grid, straight uncrossed leader lines, finer mic mesh.

GPT Image 2.6 exposes a separate complexity parameter (low, medium, high, defaulting to high) that controls how long the model plans layout geometry before painting pixels. On a 16:9 industrial teardown of a portable field recorder with six labeled aluminum modules, both low and high complexity emitted the exact same 1,002 output image tokens and settled at 0.15 credits.

The extra 18 seconds on high complexity went entirely into spatial discipline: low complexity grouped four parts down the center with right-angle bent leader lines, whereas high complexity arranged a symmetrical 3×2 grid with straight, uncrossed callouts and sharper machining ribs inside the milled chassis. Keep complexity on high for diagrams, infographics, and architectural drawings; drop it to low on single-subject edits (Tests 9–11) to shave 15 to 20 seconds off Sunburst runs for free.

Try this promptOpen in ChatCut
Overhead commercial industrial-design studio knolling photograph on smooth matte architectural grey concrete in 16:9 widescreen. Six precision-machined silver-and-orange anodized aluminum portable field-recorder components arranged in a clean exploded grid with thin crisp black technical leader lines pointing to six exact uppercase sans-serif labels printed cleanly on the surface: 'CNC CHASSIS', 'STEREO MICS', 'OLED DISPLAY', 'GAIN DIAL', 'XLR MODULE', and 'LITHIUM CELL'. Hard directional studio softbox lighting with short crisp contact shadows, tactile studio hardware aesthetic, no extra words or numbers.
Result

Test 2 · Sunburst 16:9 1K high · 0.15 credits at both low (~34s) and high (~52s) complexity.

How Do Looping GIFs Work in GPT Image 2.6? (Tests 3 & 4)

3x3 frame-by-frame contact sheet of all 9 frames from the ON AIR studio neon GIF, showing frames 1 and 9 at dim resting state with four blue bars and frames 3 through 6 flaring bright red with amber-red meter peaks.
Test 3 unpacked (frames 1–9, 1280×720 px each): frame 9 returns to frame 1's exact four-bar blue resting state so the loop never jumps.
4x2 frame-by-frame contact sheet over dark checkerboard of all 8 frames from the SIDE A cassette sticker GIF, showing spinning tape spools and two cobalt musical notes popping in frames 2–5 before clearing on frame 8.
Test 4 unpacked (frames 1–8, 960×960 px BGRA): real transparent alpha outside the white vinyl border, with notes clearing cleanly on frame 8.

The two animated loops in the opening gallery, the 16:9 “ON AIR” studio booth (t03-neon-on-air.gif) and the 1:1 “SIDE A” cassette sticker (t04-cassette-sticker.gif), were generated natively with outputFormat: "gif" at 1K on GPT Image 2.6 Flare. Under the hood, Images 2.6 renders 2 to 16 frames onto a single 1K sprite sheet, slices them apart, and encodes a looping GIF:

  • Frame count trades off against per-frame pixel dimensions. Because every frame shares one 1K sheet, our 9-frame 16:9 broadcast booth loop output at 1280×720 px (3,337 output tokens, 0.49 credits, ~24s, 1.13 MB), while our 8-frame 1:1 transparent cassette sticker output at 960×960 px (2,700 output tokens, 0.40 credits, ~23s, 1.12 MB). Any count up to 8 frames bills at the 8-frame tier, so 8 or 9 frames at 150 to 200 ms is the sweet spot.
  • State the cyclic return explicitly. In Test 3, specifying that frame 1 starts at a dim resting glow, frames 3–6 peak in bright vermilion with rising equalizer bars, and frame 9 settles back to frame 1’s exact four blue bars kept the mic boom, window mullions, and lettering locked across all nine frames.
  • Transparent GIFs write real outer alpha. Setting background: "transparent" on Test 4 produced a true BGRA matte with a crisp white kiss-cut sticker border that drops straight onto a ChatCut video track without chroma keying. For 15 more loop recipes and frame-count math, see our dedicated ChatGPT Images 2.6 GIF guide.
Try this promptOpen in ChatCut
Locked tripod close-up shot through a dark midnight recording studio glass booth window, 16:9 looping animation. Inside the dark acoustic booth, a sleek red neon sign reading 'ON AIR' above a minimal 4-bar digital LED audio level meter pulses in a clean seamless cycle: frame 1 starts at a soft dim crimson resting glow with four low blue meter bars, frames 3 to 6 flare into a bright vermilion neon halo with warm glass reflections while the LED bars jump into yellow and red peaks, and frames 7 to 9 settle smoothly back to the exact dim resting glow and four blue bars of frame 1. Condenser microphone silhouette and studio headphones in foreground stay completely still.
Result

Test 3 · Flare 16:9 1K GIF · 9 frames × 180 ms (1.62s loop) · 1280×720 px · 0.49 credits.

Try this promptOpen in ChatCut
Locked front-on 1:1 animated die-cut vinyl sticker on a fully transparent background with a crisp thick white kiss-cut border. A tilted brushed-silver and signal-orange compact cassette tape labeled 'SIDE A' in clean white sans-serif type. Across 8 frames, both white inner tape reel gears rotate smoothly while two bold electric-cobalt musical notes pop playfully out of the top-right corner on frames 2 to 5 and fade cleanly away by frame 8 so frame 8 matches frame 1's clean silhouette for a seamless loop.
Result

Test 4 · Flare 1:1 1K transparent GIF · 8 frames × 150 ms (1.20s loop) · 960×960 px BGRA · 0.40 credits.

Can GPT Image 2.6 Export Editable SVGs and Transparent PNG Stickers? (Tests 5 & 6)

Minimalist Swiss cinema club logo KINO CLUB generated as a 1K SVG on GPT Image 2.6 Flare, where the hand-outlined U and B letter paths touch at the bottom right.
Test 5A · Flare default SVG (34 lines XML, 0.55 credits): clean aperture icon, but hand-calculated <path> glyphs made U and B collide by 1 px.
Minimalist Swiss cinema club logo KINO CLUB generated as a 1K SVG on GPT Image 2.6 Sunburst using a native SVG text element with even letter spacing.
Test 5B · Sunburst SVG with <text> prompt rule (36 lines XML, 0.47 credits): even kerning, live editable copy in Figma, and 15% lower token cost.

Native vector export (outputFormat: "svg" at 1K) writes clean, human-readable XML with semantic group IDs (#film-gate, #aperture, #registration-dot, #wordmark) rather than messy auto-traced splines. Our first pass on Test 5A exposed the single most common SVG pitfall, and Test 5B proved a one-sentence fix that also lowers the bill:

  • Without guidance (Test 5A, 0.55 credits): Flare drew “KINO CLUB” as eight separate <path id="letter-..."> outlines. Hand-calculating glyph coordinates caused letter-u (ending at x=867) and letter-b (starting at x=866) to touch, and locked the brand name into uneditable curves.
  • With the <text> rule (Test 5B, 0.47 credits): Asking Sunburst for a single centered <text> element (text-anchor="middle") fixed the kerning, kept the wordmark live-editable in Figma or Illustrator, and settled 15% below the quote (0.47 vs 0.66 estimated credits) because one <text> node emits fewer XML tokens than eight outlined glyphs:
<g id="wordmark">
  <text id="kino-club-text" x="528" y="924" text-anchor="middle" fill="#070707"
    font-family="Helvetica Neue, Helvetica, Arial, Univers, sans-serif"
    font-size="126" font-weight="700" letter-spacing="10">KINO CLUB</text>
</g>
Try this promptOpen in ChatCut
Minimalist Swiss modernist cinema club logo on a warm off-white (#FAF9F6) square canvas. Above center, a bold black (#070707) horizontal 35mm film gate rectangle with four rounded sprocket holes on each side, framing a circular electric cobalt-blue (#0047FF) six-blade camera lens aperture and a vibrant vermilion-red (#FF2B2B) circular registration dot above the top-right corner. Centered cleanly beneath the emblem, write the uppercase wordmark 'KINO CLUB' as a single editable SVG <text> element with text-anchor='middle', crisp geometric grotesque sans-serif font stack, bold weight, and clean letter-spacing rather than individual letter <path> outlines. Clean semantic SVG groups and IDs, razor-sharp flat vector geometry, no gradients.
Result

Test 5B · Sunburst 1:1 1K SVG · 36 lines of semantic XML · 0.47 settled credits (quoted 0.66).

Raw transparent RGBA sticker sheet of four tactile filmmaker gear stickers with crisp white vinyl kiss-cut borders: a 35mm rangefinder camera, cobalt studio headphones, a SCENE 01 clapperboard, and a red REC tally pill.
Test 6B · Raw RGBA sticker pack PNG exported from GPT Image 2.6 Flare (1:1, 1056×1056 px, 0.27 credits) after adding our unbranded gear prompt rule.

For photographic sticker overlays rather than flat geometric marks, Test 6 (background: "transparent", outputFormat: "png" on Flare at 1K high, 0.27 credits, ~15s) delivered real studio macro textures (pebbled camera leatherette, spun-aluminum earcups, scuffed yellow clapper stripes, and glowing acrylic red plastic) wrapped in white kiss-cut vinyl borders. Two details from our run are worth copying:

  • Specify “unbranded with no manufacturer logos” on cameras, headphones, and electronics. Our first pass at 0.27 credits invented real-world camera and audio brand marks on the lens bezel and earcup; adding “completely unbranded with no manufacturer logos” and specifying “35MM 1:2” on the lens ring cleaned both up on the next 0.27-credit run.
  • Ask for transparent cutouts inside enclosed loops. Naming “a clean transparent cutout inside the headband arch” prevented opaque white fill from trapping the space under the headphone band and inside the open clapperboard jaw (visible in the dark checkerboard proof in our opening gallery).

How Accurate Is Multi-Line and Bilingual Chinese Typography? (Tests 7 & 8)

Vertical 3:4 Swiss Kunsthalle architectural photography exhibition poster showing an upward view of a brutalist concrete and glass spiral staircase against an ultramarine sky above crisp grid typography reading GLASS & LIGHT, ARCHITECTURAL PHOTOGRAPHY ZURICH, and KUNSTHALLE NOV 14 – JAN 20 CHF 16.
Test 7 · Three-tier exhibition poster on GPT Image 2.6 Flare (3:4, 912×1216 px, 1K high, ~15s, 0.20 credits): zero character errors across headlines, interpuncts, date range, and price.
Bilingual luxury fragrance lookbook card leaning beside a frosted celadon perfume bottle on travertine stone, generated on Sunburst at high quality with crisp English and Simplified Chinese text.
Test 8A · Sunburst · quality: high (~46s, 1,453 output tokens, 0.22 credits): all 13 Simplified Chinese characters and Latin serifs rendered accurately.
Bilingual luxury fragrance lookbook card leaning beside a frosted celadon perfume bottle on travertine stone, generated on Sunburst at xhigh quality with deeper travertine stone joints and finer letterpress cotton fiber.
Test 8B · Sunburst · quality: xhigh (~74s, 2,582 output tokens, 0.38 credits): deeper stone slab joints, sculpted walnut cap neck, and finer cotton letterpress deboss.

Typography is where GPT Image 2.6 pulls clearest ahead of earlier diffusion generators:

  • Three-tier Latin posters run cleanly on fast Flare (Test 7). At 3:4 1K high (0.20 credits, ~15s), Flare nailed “GLASS & LIGHT”, “ARCHITECTURAL PHOTOGRAPHY · ZURICH”, and “KUNSTHALLE · NOV 14 – JAN 20 · CHF 16”, including every middle dot, en-dash, and currency figure, on the first take.
  • Dense Simplified Chinese strokes hold at 1K high on Sunburst, while xhigh upgrades material micro-texture (Test 8). We tested 13 Simplified Chinese characters alongside English serif copy across three hierarchy sizes (“KYOTO RAIN · 京都朝露”, “HINOKI & WHITE TEA · 扁柏与白茶”, and “AUTUMN EDITION · 秋季限定 · 50 ML”). Zoomed to 100%, high quality (0.22 credits) already drew every stroke cleanly, including the 21-stroke character 露 and small-point 秋季限定. Stepping up to xhigh (0.38 credits, +73% cost) expanded output tokens from 1,453 to 2,582, buying deeper travertine wall joints and cotton-rag paper deboss rather than fixing broken glyphs, though you should still zoom-check dense 12pt CJK text on multi-row menus before publishing. For more layout templates, see our GPT Image 2 prompt patterns.
Try this promptOpen in ChatCut
Contemporary Swiss Kunsthalle architectural photography exhibition poster on heavyweight matte archival paper, 3:4 vertical format. Upper two-thirds: a dramatic upward medium-format architectural photograph of a sculptural brutalist concrete-and-glass spiral staircase flooded with hard noon sunlight against a deep ultramarine cobalt sky (#0A38A8), razor-sharp geometric shadows. Lower third: a strict Swiss International Typographic Style grid on bone-white paper (#F4F1EA) divided by thin hairline black rules and a single signal-red (#E12D23) square accent block. Exact crisp black grotesque typography in three clean tiers: headline 'GLASS & LIGHT', subhead 'ARCHITECTURAL PHOTOGRAPHY · ZURICH', and footer 'KUNSTHALLE · NOV 14 – JAN 20 · CHF 16'. Museum print quality, authentic paper grain, zero decorative clutter.
Result

Test 7 · Flare 3:4 1K high · ~15s · 0.20 settled credits.

Try this promptOpen in ChatCut
Luxury architectural fragrance campaign photograph shot on medium format, 4:5 vertical. Morning sunlight casts crisp architectural window blind shadows across a honed travertine stone plinth. On the left stands an unbranded frosted celadon-green heavy glass perfume flacon with a turned walnut wood cap and fine water droplets beside a fresh green hinoki cypress sprig. Leaning upright on the right is a thick cream cotton-rag letterpress lookbook card printed in razor-sharp charcoal ink with three exact bilingual lines: title 'KYOTO RAIN · 京都朝露', middle line 'HINOKI & WHITE TEA · 扁柏与白茶', and bottom line 'AUTUMN EDITION · 秋季限定 · 50 ML'. Every Chinese character stroke and Latin serif letter is crisp and legible, natural editorial daylight, zero yellow cast.
Result

Test 8 · Sunburst 4:5 1K · 0.22 credits at high (~46s) · 0.38 credits at xhigh (~74s).

How Steady Are Reference Edits, Product Swaps, Characters, and Sketch Renders? (Tests 9–12)

Surgical Single-Detail Photo Edit + Video Timeline Handoff (Test 9)

Reference 35mm golden-hour coastal photograph of two young women laughing while sitting on the hood of a pale slate-blue vintage convertible.
Test 9 reference input (from ChatCut's GPT Image 2.6 model gallery): pale slate-blue hood, intricate windblown hair, and warm coastal lens flare.
Edited photograph where only the vintage convertible body panels are repainted deep glossy British racing green while both women's faces, hair strands, outfits, chrome headlight bezels, and golden-hour flare remain unchanged.
Test 9 output · Sunburst (16:9, 1K high, low complexity, ~30s, 0.21 credits): deep British racing green paint with faces, hair, and fender sunstar locked.
Demo
Why reference-locked stills matter in ChatCut: once your hero frame is approved, you hand that exact frame directly to Kling or Seedance on the video timeline.

Attaching a reference photo adds input-image tokens on top of the base output quote. On Test 9, our 16:9 output quote was 0.15 credits, and the single reference photo added 0.06 credits (1,633 input tokens vs ~167 text-only), settling at 0.21 credits in 30 seconds on Sunburst with complexity: "low". Naming both the exact target (“ONLY the painted exterior body panels … to deep glossy British racing green”) and the explicit keep-list locked every windblown hair strand, freckle, hoop earring, denim cuff, and even the golden sunstar glint on the left fender.

Try this promptOpen in ChatCut
Edit ONLY the painted exterior body panels of the vintage convertible car from pale slate-blue to deep glossy British racing green, preserving natural golden-hour sunset reflections and hood highlights across the curved front fenders and hood. Keep both women sitting on the front hood 100% identical to the reference photo—their faces, smiles, freckles, windblown hair, white linen shirt and rolled jeans, rust floral sundress, chrome headlight bezels, round chrome side mirror, windshield frame, coastal cliff road, ocean horizon, and warm sun flare untouched.
Result

Test 9 · Sunburst 16:9 1K high, low complexity + 1 reference image · ~30s · 0.21 settled credits (0.15 output + 0.06 reference input).

Studio Packshot to Dark Basalt Location Campaign (Test 10)

Neutral grey studio packshot of a heavy vertical-fluted crystal perfume bottle with a brushed chrome cap and matte-black label reading NOIR VÉTIVER EAU DE PARFUM 100 ML.
Test 10A · Clean studio packshot generated on Flare (4:5, 1K high, ~16s, 0.22 credits).
The exact same fluted NOIR VÉTIVER crystal bottle relocated onto wet black volcanic basalt under a diagonal amber optical spotlight with golden caustic light rays, vetiver roots, and eucalyptus.
Test 10B · Relocated on Sunburst (4:5, 1K high, low complexity, ~38s, 0.26 credits): identical flutes, cap, and French accent mark plus physically believable glass caustics.

Relocating packaged goods fails whenever the model redraws bottle proportions or garbles small label copy. Pairing a fast Flare packshot (0.22 credits) with a Sunburst reference pass at low complexity (0.26 credits, including 0.05 credits for reference input tokens) preserved the exact flute count, brushed-chrome cap ratio, liquid meniscus, and “NOIR VÉTIVER” / “EAU DE PARFUM · 100 ML” typography, while projecting warm optical-snoot caustics through those vertical glass flutes onto wet volcanic stone.

Try this promptOpen in ChatCut
Place the exact fluted clear crystal perfume flacon from the reference photo onto a rain-wet black volcanic basalt stone slab inside a moody dark architectural studio. A narrow warm amber optical snoot spotlight beam cuts diagonally across the bottle, igniting golden caustic refractions through the vertical crystal flutes onto the wet black stone beside dark vetiver roots and misted silver-green eucalyptus leaves. Preserve the bottle's exact fluted silhouette, brushed-chrome cylindrical cap, amber liquid level, and matte-black label reading 'NOIR VÉTIVER' and 'EAU DE PARFUM · 100 ML' 100% identically without altering a single letter.
Result

Test 10B · Sunburst 4:5 1K high, low complexity + 1 reference image · ~38s · 0.26 settled credits.

Keeping a Character Consistent Across Lighting and Aspect Ratios (Test 11)

35mm daylight editorial portrait of an East Asian female art director with a chin-length black bob, straight micro-bangs, silver oval wire-rim glasses, a small beauty mark under her left eye, and an electric cobalt snap-button workwear jacket against concrete.
Test 11A · Anchor portrait on Flare (4:5, 1K high, ~15s, 0.22 credits): four distinct visual anchors established in hard afternoon daylight.
16:9 widescreen 35mm cinema still of the exact same woman in her cobalt workwear jacket and oval wire-rim glasses flipping through vinyl records inside a rainy late-night Tokyo jazz bar with glowing tube amplifiers and cyan katakana neon.
Test 11B · Cross-aspect scene transfer on Sunburst (16:9, 1K high, low complexity, ~35s, 0.20 credits): identical face, glasses, micro-bangs, and jacket under tungsten + neon light.

Jumping from a 4:5 sunlit museum courtyard to a 16:9 rainy midnight Tokyo vinyl bar stresses both facial identity and wardrobe color cast. Re-stating four concrete identity anchors from the reference portrait (chin-length bob with micro-bangs, thin silver oval wire-rim glasses, beauty mark below the left eye, and cobalt snap-button twill jacket over a white crew-neck tee) kept the character unmistakable at 0.20 credits (0.15 output + 0.05 reference input), while naturally re-lighting her skin and cobalt twill under warm vacuum-tube tungsten and cyan window neon (ジャズ レコード).

Try this promptOpen in ChatCut
Full-bleed 35mm cinema still frame in 16:9 widescreen (no film border) featuring the exact same East Asian woman from the reference portrait—preserving her identical facial bone structure, sharp chin-length jet-black bob with straight micro-bangs, thin silver oval wire-rim glasses, small beauty mark below the outer corner of her left eye, and electric cobalt-blue cotton workwear jacket with silver snap buttons over a white crew-neck tee. She is now inside an intimate late-night Tokyo jazz vinyl listening bar, three-quarter medium shot looking toward camera as her fingers flip through LP sleeves in a walnut record bin. Warm tungsten tube-amplifier and shelf glow on wooden acoustic slats, cool cyan neon reflections on rain-streaked glass behind her, natural 35mm film grain.
Result

Test 11B · Sunburst 16:9 1K high, low complexity + 1 reference image · ~35s · 0.20 settled credits.

Turning an Architectural Ink Sketch into a Medium-Format Photograph (Test 12)

Clean two-point perspective black-and-white architectural ink sketch on grid paper showing a cantilevered cliffside pavilion, three glass bays with a single pendant lamp over a sofa, left infinity pool, two right loungers, two cypress trees, and twin island peaks under a low sun.
Test 12 input · Two-point architectural ink sketch (16:9) establishing exact horizon, cantilever strut, pool edge, loungers, cypresses, and interior lamp.
Golden-hour medium-format architectural photograph matching every element of the sketch in board-formed concrete, low-iron glass, turquoise pool water, teak loungers, limestone sea cliff, and sunset sea horizon.
Test 12 output · Sunburst (16:9, 1K high, high complexity, ~54s, 0.21 credits): 1:1 structural alignment translated into board-formed concrete and golden-hour glass.

Unlike single-detail retouches where low complexity is faster, sketch-to-photo conversion benefits from complexity: "high" on Sunburst. At 0.21 credits (~54s), every structural line in our drawing (the twin island peaks beneath the low sun disk, the two cypresses at the left corner, the three glass bays with a single glowing dome pendant above a low linen sofa, the cantilevered glass-edge pool on the left, the two teak loungers on the right, and the angled concrete soffit strut meeting the limestone cliff) transferred into medium-format photography with zero leftover ink strokes.

Try this promptOpen in ChatCut
Transform this architectural ink perspective sketch into an Architectural Digest golden-hour medium-format photograph shot on Hasselblad X2D 100C, preserving the exact camera angle and composition of every sketched element: cantilevered architectural board-formed concrete pavilion with a vertical right-side concrete wall, three frameless floor-to-ceiling glass bays with a single warm glowing dome pendant lamp suspended above a low cream-linen sofa in the center bay, cantilevered turquoise infinity pool on the left terrace deck, two teak loungers on the right terrace deck, two slender Mediterranean cypress trees standing just left of the pavilion, rugged coastal limestone cliff face below the angled concrete cantilever soffit, and twin island mountain peaks beneath the low golden sun over the calm left sea horizon. Crisp architectural shadows, real concrete tie-rod texture, natural glass reflections, zero illustration lines remaining.
Result

Test 12 · Sunburst 16:9 1K high, high complexity + 1 sketch reference · ~54s · 0.21 settled credits.

All 12 Tests Compared: Credits, Timing, and When to Switch Models

Across all 24 runs in our Round 2 test suite, total settled spend was 6.30 credits, confirming that you can prototype an entire campaign’s stills, stickers, brand vectors, and looping GIFs for a few credits before spending video generation budget.

TestCapability testedBest variant & key settingsTimeOutput tokensSettled creditsKey takeaway
1Strobe beverage ad (Flare vs Sunburst × 3 + low)Sunburst (or Flare) · 4:5 1K high21s (Flare) / 54s (Sunburst)1,454 / 1,453 (681 low)0.22 (0.11 low)Same credit price; Sunburst obeyed base-splash placement 3/3 times
26-part CNC hardware knolling (complexity low vs high)Sunburst · 16:9 1K high, high complexity34s (low) / 52s (high)1,0020.15Complexity adds 0 credits; high fixed bent lines into a 3×2 grid
3Opaque 16:9 broadcast neon looping GIFFlare · 16:9 1K GIF, 9 frames × 180 ms~24s3,3370.491280×720 px frames; prompt frame 9 to match frame 1
4Transparent 1:1 cassette sticker looping GIFFlare · 1:1 1K GIF, transparent, 8f × 150 ms~23s2,7000.40True 960×960 BGRA alpha matte ready for video tracks
5Editable 1K vector cinema club SVGSunburst · 1:1 1K SVG + <text> prompt~58s (25s Flare)1,7950.47 (0.55 <path>)Native <text> tag fixed 1 px glyph collision and cut cost 15%
6Transparent RGBA filmmaker sticker sheetFlare · 1:1 1K PNG, transparent~15s1,7960.27Specify “unbranded” + inner cutouts for clean open-loop alpha
73-tier Swiss Kunsthalle exhibition posterFlare · 3:4 1K high~15s1,3440.20Fast Flare nailed headlines, middle dots, en-dashes, and prices
8Bilingual English + Simplified Chinese cardSunburst · 4:5 1K high vs xhigh46s (high) / 74s (xhigh)1,453 / 2,5820.22 / 0.381K high nailed all 13 Chinese characters; xhigh upgraded stone/paper grain
9Surgical convertible paint edit (blue → racing green)Sunburst · 16:9 1K high, low complexity + 1 ref~30s1,002 (+1,466 in)0.21+0.06 ref input credits; faces, hair strands, and sunstar stayed locked
10Fluted fragrance bottle studio-to-basalt swapFlare packshot → Sunburst low complexity + 1 ref16s + 38s1,454 + 1,4530.22 + 0.26Preserved “NOIR VÉTIVER” label while casting golden flute caustics
11Character transfer (4:5 daylight → 16:9 rainy night)Flare portrait → Sunburst low complexity + 1 ref15s + 35s1,454 + 1,0020.22 + 0.20Four explicit identity anchors held across aspect ratio and neon lighting
12Architectural ink sketch to golden-hour photoSunburst · 16:9 1K high, high complexity + 1 ref~54s1,002 (+1,472 in)0.21High complexity kept pool, loungers, lamp, cypresses, and horizon 1:1

According to OpenAI’s Image Generation API documentation and Image Prompting guide, Images 2.6 caps GIFs and SVGs at the 1K sheet tier and handles short cyclic animations rather than multi-second camera moves. Inside ChatCut, pair each model with what it does best:

  • Pick GPT Image 2.6 Flare for fast 15–24s commercial posters, transparent PNG sticker sheets, studio packshots, and 8–9 frame looping GIFs.
  • Pick GPT Image 2.6 Sunburst at high complexity for multi-callout diagrams, bilingual CJK print cards, editable SVGs, and sketch-to-photo renders, and at low complexity for surgical reference edits, product relocations, and character scene transfers.
  • Switch to Nano Banana Pro / 2.1 or Seedream 5.5 in ChatCut when you need up to 14 simultaneous reference images, ultra-wide 1:4 or 8:1 banners, or Seedream’s multi-layer RGBA decomposition (compare them in our GPT Image 2 vs Nano Banana Pro breakdown), and hand finished hero frames to Kling or Seedance for real camera movement.
Run GPT Image 2.6 Flare and Sunburst side by side on an editable video timeline.Download DesktopDownload for macOSDownload for WindowsTry GPT Image 2.6 in ChatCut

Frequently Asked Questions

Frequently asked questions

What is the difference between GPT Image 2.6 Flare and Sunburst?
Flare is the fast default variant (15 to 22 seconds per 1K still in our tests), while Sunburst is the precision variant (30 to 58 seconds) built for strict spatial instructions, dense callout grids, bilingual CJK typography, and reference-locked photo edits. In ChatCut, both variants cost the exact same credits at matching resolution and quality settings.
How much does GPT Image 2.6 cost per image, GIF, and SVG in ChatCut?
At 1K high quality without reference images, stills settled at 0.15 credits for 16:9, 0.20 credits for 3:4, 0.22 credits for 4:5, and 0.27 credits for 1:1, while low-quality drafts cost roughly half (0.11 credits on 4:5). Attaching one reference image added 0.05 to 0.06 credits in input tokens, looping GIFs settled at 0.40 to 0.49 credits, and 1K SVGs settled at 0.47 to 0.55 credits depending on XML token length.
Does the complexity setting in GPT Image 2.6 increase the credit price?
No. Raising complexity from low to high on GPT Image 2.6 emitted the exact same 1,002 output image tokens and settled at the same 0.15 credits on our 16:9 teardown test while adding about 18 seconds of pre-render layout planning. Keep complexity on high for diagrams and sketches, and drop it to low on single-detail photo edits to save wait time.
Can GPT Image 2.6 generate transparent looping GIFs and PNG stickers?
Yes. Passing transparent background with PNG output created a 1056×1056 RGBA sticker sheet at 0.27 credits, and combining transparent background with GIF output created a 960×960 8-frame BGRA animated cassette sticker at 0.40 credits that overlays directly onto video timelines without keying. See our ChatGPT Images 2.6 GIF guide for frame-count sizing rules.
How do you stop GPT Image 2.6 SVG logos from overlapping letters?
Ask explicitly in your prompt for a single centered SVG <text> element with text-anchor='middle' instead of individual letter <path> outlines. In Test 5, that one line fixed a 1-pixel U/B kerning collision, kept the wordmark live-editable in Figma, and lowered the settled charge from 0.55 credits to 0.47 credits.
Do you need xhigh quality for Chinese text in GPT Image 2.6?
Not for standard poster and card sizes. In Test 8 on Sunburst at 1K high (0.22 credits), all 13 Simplified Chinese characters (including dense 21-stroke characters like 露) rendered accurately, while xhigh (0.38 credits) mainly refined stone slab joints and cotton paper fiber texture.
When was GPT Image 2.6 released and does it include C2PA metadata?
OpenAI released ChatGPT Images 2.6 (GPT Image 2.6 Flare and Sunburst) on November 4, 2026, as documented in the Introducing Images 2.6 post and System Card. Generated outputs carry C2PA Content Credentials metadata alongside imperceptible provenance watermarking.

Checking your footage...

Less editing. More creating.

Download ChatCut Desktop free and make your first video today.

Download Desktop Download for macOS Download for Windows Try it for free