រំលងទៅមាតិកា
← កំពូល 50 នៅលើគេហទំព័រ

What's new in Higgsfield

Higgsfield · Higgsfield AI · វ៉ិប #28 · ការចំណាយ #5

  1. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Kling Image 4.0 Arrives on Higgsfield With Text-Heavy Design Support

    Kling's new image model is now available on Higgsfield. It creates images from text prompts and reference pictures. It targets realistic portraits, group photos, and designs heavy with text, like posters and product cards.

    • The model is built for natural light, skin, and material textures.
    • It can add up to 2,000 characters of text per image in 7 languages, with 70+ fonts or a font matched from a reference.
    • Prompts can run up to 15,000 tokens, enough for a detailed layout brief.
    • It can make up to 10 images in one style from a single prompt.
    អានបន្ថែម ប្រភព ↗
  2. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield Supercomputer Adds Linked Billing for Claude Through Anthropic in Beta

    Higgsfield's Supercomputer can now bill Claude API usage directly to your own Claude Platform organization through Anthropic. The feature is in beta, and image and video generation still use Higgsfield credits.

    • Claude API usage in Supercomputer can be billed by Anthropic straight to your own Claude Platform organization.
    • Linked billing is in beta, and Higgsfield is one of Anthropic's launch partners.
    • No shared keys are used; an Owner, Admin, or Billing administrator must approve the link on the Claude side.
    • Usage, monthly limits, and unlinking are managed in Claude Console.
    អានបន្ថែម ប្រភព ↗
  3. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield Adds OpenAI's GPT-6.2 Astra and Sol Models to Supercomputer

    Higgsfield has added two new OpenAI language models, GPT-6.2 Astra and GPT-6.2 Sol, to its Supercomputer tool. Astra is the higher-end option for big, demanding projects, while Sol is the cheaper choice for everyday work.

    • Astra suits template-based documents, spreadsheets, and slides that must match a template's layout, colors, and fonts, and it also writes cleaner code for websites, games, and 3D.
    • Sol handles multi-step agent tasks and natural-sounding writing, and according to OpenAI it scores within 0.5 points of Astra on FrontierCode 2.0 at about a fifth of the cost.
    • According to OpenAI, both models get better results on complex tasks while using fewer tokens (the small text units AI processes) than earlier versions.
    • To try them, open Supercomputer on Higgsfield and select GPT-6.2 Astra or GPT-6.2 Sol.
    អានបន្ថែម ប្រភព ↗
  4. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield Previews Short Drama Studio for Turning Scripts into Short Series

    Higgsfield has released an early preview of Short Drama Studio, a tool that turns a script into a short series. It splits the text into episodes, creates the characters, places, and props, and keeps the same characters across all episodes. For now it works only on desktop, and it can also be used inside Claude through Higgsfield MCP.

    • It accepts scripts in PDF, DOCX, FDX, TXT, or MD files up to 10 MB.
    • Users can choose one of six aspect ratios and 15 visual styles, including K-Drama and Anime.
    • Characters, locations, and props are generated automatically from the script text.
    • Each episode is planned into scenes and clips and can be rendered at 480p draft, 720p, or 1080p.
    អានបន្ថែម ប្រភព ↗
  5. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield Adds GPU Rental for AI Image, Video and Model Training

    Higgsfield now lets users rent a GPU (a graphics processor used for heavy AI work) right on its platform. People can use it for generating images or video, rendering, or training AI models, and a session starts in about two minutes.

    • Users pay in credits per minute while the GPU runs, and the rate appears on the GPU card and the Run a GPU button.
    • Rented GPUs appear in a My GPUs list, and users can pick a model like Wan 2.2, MiniMax H3, LTX Video, or upload their own.
    • Available GPUs range from L40S, RTX PRO 6000, and H100 for lighter jobs to H200, B200, and B300 for bigger workloads.
    • The service works on the web, in Supercomputer, and through Higgsfield MCP, and Rubin GPUs require contacting sales.
    អានបន្ថែម ប្រភព ↗
  6. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield Launches $1M Film Contest for AI-Made Movies

    AI video company Higgsfield has opened a film contest with a $1 million prize pool. Creators make a 3 to 7 minute film on any theme, and judges and viewers pick 20 winners.

    • The top prize is $200,000, with $150,000 for second and $100,000 for third.
    • Two Audience Choice awards of $50,000 each are decided by viewer votes.
    • Films are made in Higgsfield's Cinema Studio and submitted as MP4 or MOV files up to 4K.
    • Entrants must be 18 or older with a good-standing account, and no paid plan is needed. Solo or teams of up to four can enter.
    អានបន្ថែម ប្រភព ↗
  7. តម្លៃ និងដែនកំណត់ · · ផ្លូវការ · Higgsfield changelog

    Higgsfield Cuts Annual Plan Prices by Up to 60% for a Limited Time

    Higgsfield, an AI video creation service, is offering limited-time discounts on its plans. Annual plans are up to 60% off, and monthly plans get a discount for the first month.

    • The annual plan discount applies only to the first year, then the plan renews automatically at the regular price.
    • The monthly plan discount applies only to the first month.
    • Plus, Ultra, Team, and Scale plans are on sale, while Starter and Basic are excluded.
    • Offers may change, so current prices should be checked on the Pricing page.
    អានបន្ថែម ប្រភព ↗
  8. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Seedance 2.5 Adds 4K Resolution on Higgsfield

    Higgsfield's Seedance 2.5 video generator now lets users choose 4K quality, added alongside 480p, 720p, and 1080p. It works for standard generation, video extension, and draft mode.

    • 4K has been added to the quality setting.
    • Video length is 4 to 30 seconds, the same as at other resolutions.
    • A bitrate (data transfer rate) setting next to quality affects detail and cost.
    • It is not available in the US for now.
    អានបន្ថែម ប្រភព ↗
  9. ម៉ូដែលថ្មី · · ផ្លូវការ · Higgsfield changelog

    Google Launches Gemini 4 Carbon, a Lower-Cost Model for Coding and Long Tasks

    Google's new AI model, Gemini 4 Carbon, is now available in Higgsfield's Supercomputer. It succeeds Gemini 4 Argon and carries a low-cost label in the model selector.

    • It is well suited to coding, building websites, apps, and games from start to finish.
    • It can handle multi-step tasks where the AI plans the order of the work itself.
    • Google says it performs well on cybersecurity tasks.
    • To try it, open Supercomputer and select Gemini 4 Carbon.
    អានបន្ថែម ប្រភព ↗
  10. ម៉ូដែលថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield Adds Claude Fable 5.2 for Large-Scale Work

    Higgsfield's Supercomputer now offers Anthropic's new model, Claude Fable 5.2. It is built for large jobs such as full websites and games, 3D scenes, complex code, and long analysis of many documents or data sets.

    • Compared with Claude Fable 5.1, it better grasps the intent behind vague requests and carries large projects through.
    • It keeps refining complex games and 3D builds instead of stalling.
    • It reads charts and pictures more accurately and finds mismatches across many documents.
    • Anthropic says responses are up to 30% faster and explanations are clearer.
    អានបន្ថែម ប្រភព ↗
  11. ម៉ូដែលថ្មី · · ផ្លូវការ · Higgsfield changelog

    ByteDance Releases Seedream 5.5 Image Model With Transparent Backgrounds on Higgsfield

    ByteDance has released Seedream 5.5, a new image generation model, now available on the Higgsfield platform. It is said to improve on the previous Seedream 5.0 in several areas.

    • You can set the background to transparent when generating, so you get a PNG with only the subject.
    • Multiple rounds of edits stack up, and parts you did not change stay as they were.
    • Small and multilingual text comes out legible on labels, packaging, signs, and posters.
    • Output comes in 1K, 1.5K, or 2K resolution, with nine aspect ratios from 1:1 to 21:9.
    អានបន្ថែម ប្រភព ↗
  12. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield's Ads Studio now makes long video ads through chat

    Higgsfield's Ads Studio lets users make video ads by chatting with an AI agent (an AI that splits work into steps on its own). Users can enter an idea or script, then pick a style, product, and actor, and the tool can make even long ads.

    • You can type an idea or paste a script into the chat to start an ad.
    • Video styles include micro drama, cartoon story, and cinematic story, while image ads use Meta ad styles.
    • Entering a website address fills in a brand kit with colors, selling points, and products automatically.
    • For each ad, you can choose an actor from the library, an AI influencer, or a newly created avatar.
    អានបន្ថែម ប្រភព ↗
  13. ម៉ូដែលថ្មី · · ផ្លូវការ · Higgsfield changelog

    Thinking Machines' Inkling 2 Small Now Available on Higgsfield Supercomputer

    Thinking Machines Lab's new AI model, Inkling 2 Small, is now available in Higgsfield's Supercomputer. It is built for long, multi-step work such as coding, building websites, 3D, and documents.

    • It can read and understand screenshots, charts, and text in images.
    • It can search documents, work with spreadsheets, and produce finished reports, with finance as its strongest area.
    • Its effort level can be adjusted from faster to more thorough, and it is set to Medium by default.
    • For quick questions and simple chat, other models in the selector are a better fit.
    អានបន្ថែម ប្រភព ↗
  14. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Google Releases Gemini Omni Flash 1.2 Video Model With Native Audio

    Google has released Gemini Omni Flash 1.2, a new version of its video model. It creates clips with built-in sound and can extend an existing video. Higgsfield has made the model available on its own platform.

    • You can make videos from text, images, or video.
    • Each generation runs 3 to 15 seconds, at resolutions from 360p up to 4K.
    • You can use up to 6 reference images and up to 3 short reference videos.
    • You can extend a video from its end in 3 to 10 second steps, up to 30 seconds total.
    អានបន្ថែម ប្រភព ↗
  15. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield Adds GPT Image 2.6 in Flare and Sunburst Versions

    Higgsfield now offers OpenAI's new image model, GPT Image 2.6. Besides regular images, it can make animated GIFs, vector SVGs, and images with transparent backgrounds.

    • There are two versions: Flare is the default, suited to everyday content, drafts, and image series.
    • Sunburst is for detailed edits and final layouts where faces, products, and text must stay close to the original, but it takes longer to generate.
    • GIFs turn one scene and one simple movement into a loop, with 2 to 64 frames depending on frame size.
    អានបន្ថែម ប្រភព ↗
  16. ម៉ូដែលថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield adds Kling 4.0 Flash, a lighter model for quick short videos

    AI video service Higgsfield now offers Kling 4.0 Flash, a lighter version of Kling 4.0. It is built for quick tests and frequent short videos, and each generation costs less than with Kling 4.0.

    • Videos run 3 to 20 seconds at 720p.
    • It keeps the References, Frames and Edit video modes but offers fewer controls.
    • Users can set only the opening frame; ending frames and keyframes (the anchor shots of a scene) are not available.
    • It accepts up to 10 reference images and 5 reference videos.
    អានបន្ថែម ប្រភព ↗
  17. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield launches Shorts Studio for vertical videos hosted by AI avatars

    Higgsfield released Shorts Studio, a tool for making 9:16 vertical short videos with an AI avatar as the host. Add a character, set a topic and upload a reference video for the motion, and the video is generated.

    • It writes the script for you from a topic, an attention-grabbing opening line (hook) or a few key points.
    • Motion comes from a reference video 10 to 120 seconds long or a ready-made format in Explore.
    • Length is adjustable from 10 seconds to 2 minutes, with the credit price depending on duration.
    • It works on Higgsfield or inside Claude through Higgsfield MCP (a way to connect outside tools to AI assistants).
    អានបន្ថែម ប្រភព ↗
  18. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield launches Higgsfield Mangaka for making manga and webtoons

    Higgsfield released Higgsfield Mangaka, a new tool for making manga, comics and webtoons. Users start from an idea or a script, create characters, generate chapters and pages, edit them and export the finished book.

    • Users can pick black-and-white or color manga, comics or webtoon styles, an aspect ratio and one of 14 languages.
    • In Auto mode the tool builds the cast, settings, props and chapters by itself, while in Manual mode users create and review each piece.
    • The Recreate feature redraws an existing manga with its characters swapped for people from your own photos.
    • Books of 4 to 30 chapters can be exported as PDF, CBZ (a comic book file format), PNG or webtoon format, or published on Higgsfield.
    អានបន្ថែម ប្រភព ↗
  19. មុខងារថ្មី · · ផ្លូវការ · Higgsfield changelog

    Higgsfield Intone launches for lip-synced talking character videos

    Higgsfield released Higgsfield Intone, a tool for making videos of talking characters. Pick a character and a location and add speech, and you get a lip-synced video with a consistent voice across scenes.

    • Characters can come from an uploaded image, including your own photo, or from AI Influencer.
    • Speech can be an uploaded recording or newly generated audio, up to 30 seconds per generation.
    • Bracketed cues like [smirks] or [short pause] let users direct how each line is spoken.
    • Users can choose 720p or 1080p and one of five aspect ratios, and set movement and camera with an optional prompt (text instruction).
    អានបន្ថែម ប្រភព ↗

AI អានការបង្ហោះផ្លូវការ និងសារព័ត៌មានពាណិជ្ជកម្ម ហើយសរសេរវាឡើងវិញជាពាក្យសាមញ្ញ។ ធាតុមួយចំនួនអាចនឹងបាត់ ឬខុស។ សូមពិនិត្យមើលប្រភពសម្រាប់ព័ត៌មានលម្អិត។

អ្នក