HeyGen Voices now supports ElevenLabs v4, which the company calls ElevenLabs' most expressive text-to-speech model to date. v4 delivers script readings closer to voice actor delivery, maintaining rhythm and tone consistently from start to finish on long scripts, and preserving the same voice timbre when regenerating the same line. To use it, select a voice in Voices and choose ElevenLabs v4.
OpenAI removed DALL-E 3 from the API on May 12, 2026, and image generation built into the free version of ChatGPT is limited to about 2 to 3 images per day.
Topaz AI is not free: Topaz Labs ended perpetual licenses for Topaz Photo AI and Topaz Video AI in September 2025, and the desktop versions no longer offer trials, switching to pay-first with refunds available within 48 hours.
LetsEnhance is a browser-based AI upscaling tool run by a Ukrainian team, with a 4.1 score from 186 Trustpilot reviews; the paid version supports up to 16x upscaling and 512 MP output, while free accounts get only 10 credits and watermarked output.
InvokeAI released 7.0.0 alpha 1, calling it its largest update ever, with about 320 PRs and 2,600 commits, requiring Python 3.12, and with the new interface as default.
Why it matters: InvokeAI 7 integrates canvas, workflows, and video generation into a single workbench, and adds support for models such as MiniMax H3 and LTX-2.5, which can be used to judge whether to migrate an existing image generation workflow.
Krita AI Diffusion released the 1.54.0-pre prerelease, which has migrated to Krita 6 / Qt6 and is no longer compatible with Krita 5.x; Krita 6.0.4 is recommended. This version adds support for Qwen Image 2.1; the official notice states this is a prerelease that has not been fully tested and performs poorly on macOS.
FastVideo released the experimental model FastH3 Trim, which removes 8 of the least impactful blocks from FastH3 V2's 50 MiniMax H3 transformer blocks and replaces each block's AdaLN timestep projection with a shared rank-16 basis, trained with 8-step DMD2 (checkpoint 300), enabling 8-step text-to-video generation with synchronized audio.
Why it matters: FastH3 Trim uses pruning plus shared AdaLN to bring 8-step audio-video generation to consumer-grade hardware, and can be compared against V2 to weigh image quality against size.
IndexTeam has released Index-Translate-35B-A3B-preview, a multilingual text translation model based on Qwen3.5 with 35B total parameters and 3B active parameters in an MoE architecture. It supports 150 languages and natively supports end-to-end image translation without requiring external OCR.
Why it matters: The Index-Translate family adds a 35B-A3B preview version supporting 150 languages and direct image translation, allowing evaluation of its low-resource language performance against report data.
Google released its frontier model Gemini 4 Argon in September, featuring a 1 million token output limit and focusing on advanced reasoning and cybersecurity defense. It is currently being rolled out in phases to trusted cybersecurity professionals through the Fairwind Program.
LTX-2 has released v1.4.2, adding the ltx_pipelines.alpha_gen pipeline. This pipeline is a single-stage IC-LoRA process based on the full model, using video CFG 1 with STG disabled.
Black Forest Labs' FLUX 3 Image has joined the Envato AI model lineup, supporting multi-reference image generation and editing, native 2K/4K high-resolution output, in-image text generation and modification, reference-based style transfer, and precise local editing. Users can specify the exact role of each reference image and add, remove or replace specific elements while keeping the rest of the image unchanged.
Midjourney launched sidebar style previews, larger Create images and folder default parameter settings for alpha.midjourney.com; this week focused mainly on bug fixes and cleanup. Multiple srefs in the prompt bar are now merged into a single style pill with a thumbnail, hovering shows the code, and --exp can be pinned as a folder default. The company said new collaboration tools will launch next week and teased persistent edit history and better ways to explore styles.
WaveSpeedAI released Ultimate Video Upscaler, which can upscale video to 720p, 1080p, 2K or 4K; the official example upscales a 480p 5-second clip to 3840 × 2160 while keeping 24 fps and the original audio.
Why it matters: A workflow of low-resolution drafts plus post-upscaling, combined with a cost structure priced by target resolution, can be used to evaluate trade-offs in the production pipeline.
Kling 4.0 has entered closed beta, with an official release planned for October this year; upgrades include stable dynamic motion, high-quality stereo audio, more accurate lip sync, up to 4K resolution and 10-bit HDR output.
Why it matters: Kling 4.0 enters closed beta; creators test 30-second long takes, motion stability and multi-shot consistency, which can help judge whether a commercial production pipeline is worth switching to.
WaveSpeedAI launched AI Cinema Studio in the generator sidebar, using five cards—References, Film setup, Camera, Color palette and Lighting—to turn cinematic style, camera model, camera movement and lighting choices into precise shot descriptions for video models.
Why it matters: Cinematic feel is broken down into preset panels for style, camera model, camera movement and more, letting creators judge whether this saves them the step of writing prompts.
Sony Pictures Imageworks texture lead Jody Tidsbury and motion graphics artist Nori Kaneko disclosed that the team used Adobe Substance 3D Painter, Designer, Sampler and After Effects to complete the character, costume and environment textures and the concert motion graphics for "KPop Demon Hunters."
The official Comfy blog shares YUI, the first animated short created by animation creator 852 Hakoniwa using Comfy Agent, whose protagonist's name is taken from the last three letters of ComfyUI.
Why it matters: The animation creator completed a first short with Comfy Agent and disclosed the cost, work hours, and a shot-by-shot approach of comparing models.
OpenAI explores how the greatest value of advanced AI may lie in the everyday execution work behind breakthrough ideas rather than in the ideas themselves. The article argues that the execution stage will shape the next phase of the economy and the pace of progress.
Suno released Speech beta, its first audio model that generates human voice and music together as a single complete track. Users input an idea, poem, or existing text, then describe the desired vocal tone and music style, and the system generates voice content with original background music. The model is now open to all users in beta; it had previously been tested with a small group of users for one month. The official statement notes that British accents occasionally drift toward Australian, and other effects still need improvement.
Why it matters: Suno synthesizes voice and background music into a single track, allowing creators who do short-drama narration or dubbing to evaluate a new production approach.
Luma introduces Variants: after a user uploads an approved static ad, they select a layout and language, and the system automatically generates versions for each size and language combination, keeping the logo, headline, and CTA unchanged while only adjusting layout and copy.
Why it matters: Ad teams can use this to evaluate batch-adapting a single approved asset into multiple sizes and language versions, replacing manual rework.
Runway released Runway Ads, a product that runs and optimizes the creative side of paid advertising end to end: after connecting ad accounts and brand assets, it can generate video and image creatives, publish approved variants directly to Meta, Google, and TikTok, then read delivery performance from these platforms and generate the next round of creatives accordingly.
Why it matters: Runway connects ad creative generation, delivery, and performance feedback into a closed loop, and provides conversion and cost data from its own ad delivery, which can be used to judge the usable boundaries of automated ad delivery tools.
facefusion released version v4-beta, adding a Timestamp alias type. This version was tagged by henryruhs on September 30, and specific feature details have not yet been elaborated in the release notes.
LTX-2 released v1.4.1, modifying the HDR video encoding process to output decoded data blocks directly as EXR and HLG, no longer retaining the full clip in memory. This version was released by lt-research-bot on September 30 and includes 1 commit to the main branch.
facefusion released version 3.9.1, updating to the latest onnxruntime version and limiting the arena workaround to affected onnxruntime versions. This version also fixes a performance issue with the content analyzer under CoreML.
Shaanxi micro-dramas "Echoes of Mountains and Rivers" and "Song of Changle" were selected for the "Annual Micro-Drama" list of the micro-drama honor unit established for the first time at the 16th China Golden Eagle TV Art Festival. "Echoes of Mountains and Rivers" was co-produced by Hongguo Short Drama and Wubao County Hongsheng Film and Television, pioneering dual-version broadcast in vertical and horizontal formats, with over 1.2 billion views across the web; "Song of Changle" was produced by Xiying Video and others, using on-location filming at heritage sites and AIGC digital technology to restore historical scenes, and its horizontal version has aired on CCTV-6.
The General Office of the National Radio and Television Administration has launched the 2026 International Micro-Drama Competition under the theme "Micro Connects the World, Drama Illuminates a New Journey," hosted by the Sichuan Provincial Radio and Television Administration and organized by Chengdu Radio and Television Station, with five competition units: Chinese Concepts, Asia-Pacific Harmony, Everyday Life, Civilizational Coexistence, and Original Screenplays. The competition solicits completed micro-drama works and original screenplays globally, with submissions due by November 30, 2026; works broadcast domestically must indicate a network audiovisual program license number or filing number.
The Shaanxi-recommended micro-drama "On the East Mountain Top" was selected in the review results of the National Radio and Television Administration's Outstanding Online Audiovisual Works Selection for Q1 2026. The drama was co-produced by Xizang Minzu University, Shaanxi Zhongqin Hongji Film and Television Communication Co., Ltd., and Shaanxi Huashang Film Co., Ltd. Adapted from Zhou Weituan's award-winning novel of the same name, it is a 2025 Shaanxi Provincial Major Cultural Premium Support Project and was included in the second batch of the Administration's "Micro-Drama+" Creation Plan recommendation catalog.
On September 30, the 2026 Premium Micro-Drama Autumn Review Meeting, guided by the Network Audiovisual Program Management Department of the National Radio and Television Administration and hosted by the China Television Arts Commission, was held in Beijing. Creative representatives of 12 micro-dramas, including "A Glance · Nanyang" "Serious Case Squad 6: The Vanished Badge Number" and "Legend of the Tang Dynasty Youth Tianxing," introduced their creative work, and attending experts offered in-depth reviews. The meeting noted that this quarter's premium micro-dramas span diverse themes with high recognizability, proving that micro-dramas can accommodate both "horizontal" and "vertical," "long" and "short," and "live-action" and "AI."
Hugging Face released the Open TTS Leaderboard, using objective metrics to evaluate open-source multilingual TTS and voice cloning models, shortening single-model evaluation time from weeks of voting to a few hours.
Kling officially released a guide to using Kling 4.0 Multiple Keyframes, supporting up to 10 keyframe images per generation, used to set visual nodes on the timeline such as character entrances, product reveals, and scene switches, with the model generating the motion and transitions between the nodes accordingly.
Why it matters: Kling officially provides a complete method for controlling 30-second long videos with multiple keyframes, including timeline planning and common mistakes, which can be directly transferred to advertising and short-narrative workflows.
The third scheduling meeting of the NRTA's micro-short-drama premium creation and dissemination plan "Five Batches Project" was held on September 28, requiring close adherence to the directions of premiumization, popularization, and internationalization, continued support for live-action dramas, and guidance for the healthy and standardized development of AI dramas. The meeting reported on recent key creation-management work; Guangxi, Henan, and Liaoning introduced premium creation guidance; Hongguo Short Drama and iQIYI introduced supported works and the broadcast of Long March-themed works. The meeting required improving creation quality, increasing broadcast efficiency, landing funds, and standardizing management, and implementing the filing review, license issuance, labeling, and other provisions of the "Measures for the Development and Management of Micro Short Dramas."
LTX-2 released v1.4.0, adding --chunked to Distilled, the two-stage TI2Vid, A2Vid, Dub-It and IC-LoRA, enabling ultra-long video generation with a default 97-frame window and 25-frame carry, without needing to set --chunk-pixel-frames or --chunk-carry-frames.
Why it matters: LTX-2 v1.4.0 lists interface changes for long-video generation, chunked decoding and HDR workflows in one place, making it easier for existing workflows to assess migration costs.
Canva has launched an AI Connector for Meta's personal AI agent Muse, allowing Muse to hand off visual creation tasks directly to Canva. Once connected, Muse can access users' brand fonts, colors, and assets, search the design library, and export finished files, with all generated designs remaining fully editable. The feature is now available to users in the United States and Canada, with AI capabilities secured by Canva Shield.
OpenAI DevDay 2026 announced over 20 releases, covering GPT-6 Astra, ChatGPT, Codex, APIs, security, and new developer tools. Specific feature details and availability have not yet been elaborated in the original text.
Kyutai fine-tuned OVIE for 50,000 steps on the DL3DV dataset and released kyutai/ovie-ft-dl3dv, which shares the same architecture as the base model with only the weights differing.
Kyutai has released OVIE-512 on Hugging Face. This is a base model from the OVIE series, retrained to 512×512 using the same recipe, trained for 250K steps with a global batch size of 256.
Kyutai has released OVIE-ft-re10k on Hugging Face. This is a checkpoint of OVIE fine-tuned for 50K steps on RealEstate10K multi-view data, accounting for only 2% of its 2-million-step pretraining budget. The architecture is identical to the base model, and it can be loaded directly by reusing the same code path.
Tomato Novel has launched Phase 3 of the 'IP Echo Plan,' soliciting short-drama and comic-drama adaptations of Tomato short stories. The announcement was published by Tomato Novel, a ByteDance subsidiary; specific adaptation rights and registration details have not been disclosed in the page body.
Tomato Novel's Short Story 'Thousand Words, Ten Thousand Gold' Incentive Plan Phase 4 has launched with a prize pool of over ¥5 million. Short-story revenue sharing remains at 10:0 (author 10). Guaranteed minimums are ¥80–500+ per thousand words for ultra-short stories and ¥100–700+ per thousand words for medium-short stories, with a maximum of ¥60,000 for a single hit. The reward pool has expanded from TOP 200 to TOP 600. New incentives include multi-book bonuses (up to an additional 25% of new media channel revenue per author) and TOP 10 incentives across seven categories. Works may also be adapted into short dramas or comic dramas, with adaptation revenue stacked on top.
Tomato Novel has launched the short-story 'IP Gold Collection Co-Creation Program,' focusing on comic drama adaptation directions and mid-to-short lengths, covering 11 popular categories including male/female-oriented fantasy, xuanhuan/xianxia, and contemporary sweet romance. Participating works must exceed 20,000 words. Editors can sign guaranteed minimum contracts for quality works; premium content receives additional manuscript fee incentives; top works in the same category can stack blockbuster chart bonuses, the Ten-Thousand-Gold-Per-Thousand-Words bonus, and the Echo Program bonus; high-quality, high-output authors receive additional bonuses.