Skip to content

I Ranked Every AI Model So You Don't Have To

2026-01-29 · long tutorial / working session · Adil · recording, 10:06 · watch on YouTube

Topic: Image generation & editing (also: Video generation & shot direction, Assets, references & consistency, Higgsfield platform & ecosystem, Prompting & AI assistants, Cost, credits & plans, Post-production & finishing) · mode: Tutorial · tags: giveaway, vendor claims

Adil gives beginners a tour of Higgsfield, covering image and video generation, editing tools, Soul ID, Cinema Studio and apps. He ranks the models by how often top creators on the site use them. Nano Banana Pro is his top image model. For video, Kling 2.6 ranks top, just above Sora 2, which does not accept human faces as references. His practical advice: test on Unlimited models, animate your own images rather than using text-to-video, train Soul ID on 20+ photos for a consistent face, and keep iterating.

What is said

The parts of the video and what is said in each (★ — how much it matters to the video).

00:00:00 — Intro: one platform for every top-tier model

  • ★ Adil presents Higgsfield as the only platform that combines every top-tier AI model under one roof. — 00:00:00 · Higgsfield AI

00:00:49 — Interface tour and image generation: aspect ratio, Unlimited models, Soul, Nano Banana Pro, Z-Image

  • ★★ In Higgsfield's image generator, Adil advises choosing 16:9 for YouTube or cinematic shots and 9:16 for vertical content on TikTok or Reels. — 00:00:49
  • ★★★ Adil's tip: use a model marked Unlimited on Higgsfield for testing so you don't burn credits. — 00:01:24 · Higgsfield Unlimited, credit
  • ★★ Adil calls Soul, Higgsfield's native image model, the go-to for ultra-realistic fashion visuals and perfect faces. — 00:01:24 · Soul
  • ★★ Adil says Nano Banana Pro has a smart LLM under the hood, so it reasons rather than only generates and follows logic better than any other image model. — 00:01:48 · Nano Banana Pro
  • ★★ Adil ranks the image and video models by how often top creators use them on Higgsfield. — 00:01:48 · Higgsfield AI, image-to-video

00:02:08 — Tier list of image models

  • ★★ In Adil's January 2026 tier list of image models, Soul ranks A tier (as heard), fantastic for ultra-realistic fashion and UGC content. — 00:02:16 · Soul
  • ★★ Adil ranks GPT Image and Nano Banana C tier among image models: good for many tasks but not exceptional. — 00:02:16 · GPT Image, Nano Banana
  • ★★ Adil ranks Seedream 4.5 A tier among image models, for great face consistency and overall great graphics. — 00:02:16 · Seedream 4.5
  • ★★ Adil ranks Flux 2 Pro B tier among image models: solid and reliable, but short of the S- and A-tier models. — 00:02:16 · Flux.2
  • ★ Adil ranks Reve D tier among image models, mostly used for graphic design tasks. — 00:02:44
  • ★★ Adil ranks Z-Image D tier: it generates fast but cannot take uploaded reference images. — 00:02:44 · Z-Image
  • ★ Adil puts two more image models in E tier for their narrow use cases (names garbled in the transcript: "Klingon 1 image" and "1.2.2 image"). — 00:02:44 · Kling
  • ★★★ Adil ranks Nano Banana Pro easily S tier among image models: great at following prompts, with deep knowledge of historical and real-life data, and better text rendering than anything else on the market. — 00:02:44 · Nano Banana Pro

00:03:10 — Image editing tools: Inpaint, Relight, Instadump, Photodump, Face Swap

  • ★★ When an image you like has a flaw such as six fingers, Adil advises not re-rolling but using Inpaint: brush over the hand and type "hand with five fingers". — 00:03:10
  • ★★ Higgsfield Relight changes the lighting colour and direction of existing shots. — 00:03:36
  • ★★ Higgsfield Instadump generates many images of you from one selfie. — 00:03:36
  • ★★ Photodump works like Instadump but takes a Soul ID as input, so its results are much more consistent. — 00:03:36 · Soul ID
  • ★★ Higgsfield Face Swap takes a face to insert and the image to put it in, and swaps the face with no prompt. — 00:03:36 · face swap

00:04:04 — Image-to-video versus text-to-video

  • ★★★ Adil calls image-to-video the number one trick beginners miss: animate an image you made; text-to-video is for quick results when you need little control, but image-to-video gives the most consistent output. — 00:04:04 · image-to-video, text-to-video

00:04:28 — Video models by job and their tier list

  • ★★ Adil says Kling 2.6 gives the cleanest-looking videos with the most control over the camera and realism. — 00:04:28 · Kling 2.6
  • ★★ Adil recommends Sora 2 for viral, short, punchy clips and calls it very realistic. — 00:04:28 · Sora 2
  • ★★★ For human emotions and speech, Adil's go-tos are Veo and Seedance (Veo 3.1 and Seedance 1.5 Pro in his ranking), which he calls the immersion kings. — 00:04:28 · Veo 3.1
  • ★★ Adil ranks Seedance 1.5 Pro B tier among video models: great at human emotions and dialogue, but it struggles to generate new environments when the camera moves. — 00:04:28
  • ★★ Adil ranks Veo 3.1 A tier among video models for great realism and audio quality. — 00:05:03 · Veo 3.1
  • ★★ Adil ranks MiniMax Hailuo 2.3 C tier: a decent video model, but nothing outstanding next to those above. — 00:05:03 · Hailuo 2.3, MiniMax
  • ★★★ Adil ranks Kling 2.6 S tier among video models, the go-to for pretty much every use case. — 00:05:03 · Kling 2.6
  • ★★ Adil ranks Wan 2.6 D tier: very narrow use cases and not very reliable. — 00:05:03 · Wan 2.6
  • ★★★ Adil ranks Sora 2 S tier, the gold standard for physics and consistency, but slightly below Kling 2.6 because it does not accept human faces as references and costs more than the other models. — 00:05:32 · Sora 2, Kling 2.6
  • ★★ Each video model on Higgsfield comes with presets that need no prompt; in the Create Video tab, click Change, pick the model at the top and the preset at the bottom. — 00:05:32

00:05:48 — Video presets, Mixed Media Studio, Kling Edit, Kling Motion Control

  • ★★ Mixed Media Studio, a Higgsfield tool for music-video and content creators to stylise clips, comes with over 20 curated presets, each fully customisable. — 00:06:01
  • ★★ With Kling Edit you upload a video and change it with reference images or text, instead of regenerating it from scratch to change a detail. — 00:06:01 · Kling O1, Kling
  • ★★★ When prompting does not give the subject movement or camera control you want, Adil advises Kling Motion Control: upload a reference video and it copies its movement. — 00:06:28 · motion transfer
  • ★★★ To keep a face consistent, upload at least 20 photos of yourself or a generated character from many angles to Soul ID; Higgsfield learns the face in minutes and puts that person in any scene. — 00:06:28 · Soul ID

00:06:48 — Soul ID and AI Influencer Studio for consistent characters

  • ★★ Higgsfield's AI Influencer Studio generates characters from scratch with every detail customisable, as in games. — 00:07:16 · AI Influencer Studio
  • ★★ In Higgsfield Cinema Studio you select specific cameras, lenses, aperture and focal length. — 00:07:16 · Cinema Studio

00:07:33 — Cinema Studio and one-click apps

  • ★★ Higgsfield's Apps tab holds one-click templates with no prompt: Transitions makes seamless transitions between two shots and Recast swaps the character in any video in seconds. — 00:07:45
  • ★★ The Shots app on Higgsfield makes nine cinematic shots from a single image. — 00:08:12
  • ★★ On Higgsfield's Explore page you can clone all the assets of another user's project and recreate it yourself. — 00:08:12 · Higgsfield AI

00:08:23 — Asset library, Explore page and Community

  • ★★ For a specific composition, Adil advises dragging in a reference image instead of relying on text alone. — 00:08:41
  • ★★ Keyframes are anchors: the start frame is where the video begins, the end frame where it must finish, and the AI generates the rest. — 00:08:41 · keyframe

00:08:46 — Rapid-fire answers: references, keyframes, iteration

  • ★★★ Adil's number one advice for beginners is iteration: without a perfect first result, tweak or simplify the prompt slightly and rerun, or fix it with the editing tools. — 00:09:10
  • ★★ Adil says that for the last 100 years creativity was limited by budget, such as a million dollars for a sci-fi movie, and that this barrier is gone. — 00:09:10

00:09:20 — Closing thought on budgets and an Ultimate plan giveaway

Threads in this video

Where the stories that moved here went next

Mentioned

Central: image-to-video · Kling 2.6 · Nano Banana Pro · Soul ID

Also: Projects: Higgsfield Community

Organizations: MiniMax

Concepts: face swap · keyframe · motion transfer · text-to-video

Products: AI Influencer Studio · Cinema Studio · Flux.2 · GPT Image · Hailuo 2.3 · Higgsfield Face Swap · Higgsfield Recast · Higgsfield Unlimited · Inpaint · Instadump · Kling · Kling Motion Control · Kling O1 · Mixed Media Studio · Nano Banana · Photodump Studio · Relight · Reve · Seedance 1.5 · Seedream 4.5 · Shots · Sora 2 · Soul · Veo 3.1 · Wan 2.6 · Z-Image

Replied to by

← Previous: Grok Imagine: Best Use Cases for AI Video Generation · Next → Ranking Every AI Video Model (S-Tier Revealed 🏆)