Choosing a model

PonPon gives you one tab and a shelf of models — eight for images, twelve for video. You don't have to learn them all. This page is a map: what each one is best at, and how to pick without overthinking it.

Match the model to the job

Pick for the thing your shot actually needs — text rendering, physics, camera control, speed — not for the brand name. Every model has one or two things it does better than the rest; choose for that and let the rest go.

Two other dimensions matter once you're past "which brand":

• Speed & cost — Fast tiers return sooner and cost fewer credits; Pro tiers cost more for higher resolution or length. The credit cost shows on the Generate button before you commit.
• Tier — most families ship a Standard and a Fast (or Pro) variant, and the prompt carries across them unchanged. Draft cheap, finish high. More below.

Image models

Open the image generator and switch models from the picker. PonPon defaults to GPT Image 2. The links below each go to a deep-dive on that model's standout capability.

• GPT Image 2 — the default and best all-rounder: strongest prompt adherence, the most legible in-image text, and generation plus in-place editing in one model. GPT Image 1.5 is the precision, true-color tier.
• Nano Banana Pro — surgical, maskless object edits, strong character and product consistency, accurate in-image text, up to 4K. Nano Banana 2 is the speed-tuned sibling for the same edits at flash speed.
• Seedream 5.0 — editorial photorealism, intelligent visual reasoning (hands, gaze, depth), and reliable text in images. Seedream 4.5 is the faster, cheaper tier.
• Midjourney V8 — the signature cinematic, painterly look, no Discord required (renders four options per generation).
• Grok Image Generator — xAI's highly aesthetic text-to-image, with editing.

Video models

Open the video generator and switch models from the picker.

• Veo 3.1 — the most controllable camera language plus native audio; the all-rounder when the move matters. Veo 3.1 Fast drafts the same look quicker.
• Sora 2 — best-in-class physics and texture realism with synced audio, up to 12-second clips. Sora 2 Pro adds longer clips, higher resolution, and a priority queue.
• Kling 3.0 — the most feature-rich: lip-sync, multi-shot storytelling, motion-brush control, native 4K, and strong image-to-video. Kling 2.6 Pro is the dependable previous generation, Kling O1 is cost-efficient, and Kling O3 is editing-focused (video-to-video and restyle).
• Seedance 2.0 — fast, expressive, vertical-first social clips with audio-visual beat sync. Seedance 2.0 Fast pushes generation speed further.
• HappyHorse — the most versatile pipeline: text, image, reference, and video-to-video editing, with many reference characters and native audio.
• Grok Imagine — xAI's text- and image-to-video with audio.

Pick by what you need

Compare in practice

The cheapest way to choose is to run one prompt on two or three models and keep the best take. Take a single brief:

A barista latte-arts a heart, slow push-in, warm morning light. 9:16, 5 seconds.

• On Veo 3.1 the camera push reads cleanly and the pour syncs with subtle ambient sound.
• On Sora 2 the milk and crema behave most convincingly — physics carries the shot.
• On Seedance 2.0 you get a punchy, vertical-native take fastest and cheapest.

Same words, three strengths. You learn more from one side-by-side than from any spec sheet.

Head-to-head comparisons

When two models are genuinely close, a direct comparison settles it:

• Sora 2 vs Veo 3.1 — physics realism vs the most precise camera control and audio.
• Kling 3.0 vs Sora 2 — dialogue and multi-shot storytelling vs world-accurate physics.
• Nano Banana Pro vs Seedream 5.0 — surgical, maskless editing vs editorial photorealism.
• Nano Banana Pro vs Midjourney V8 — precise editing and accurate text vs the cinematic, painterly look.

Standard, Fast, and Pro tiers

Several families ship more than one tier, and the prompt carries across them unchanged:

• Fast tiers — Veo 3.1 Fast, Seedance 2.0 Fast, Nano Banana 2, Seedream 4.5 — trade a little fidelity for speed and lower cost, ideal while you're still iterating.
• Pro tiers — Sora 2 Pro — add resolution, length, or queue priority for the final render.

Some jobs are a tool, not a model

A few choices aren't a model decision at all — they're a dedicated tool:

• Portraits and fashion — switch the image picker to Muse for a guided character pipeline.
• Background removal, upscaling, angle changes, text fixes — remove background, upscale, multi-angle, and text edit.
• One-tap themed videos — the Effects library picks the model and prompt for you.

Ready to put a model to work? Start with Text-to-video basics or Image generation basics.