Visual
Image, video and suites-
Live
Kling
KuaishouKling 4.0Official Kling app. Kling 4.0 has been open to everyone since 15 Oct 2026: up to 30 seconds in one pass, Omni Reference with up to 15 references and 10 keyframes. Since 28 Oct 2026 the Kling 4.0 API is fully available, with text to video, image to video and Omni video at 720p and 1080p and new 21:9 and adaptive aspect ratios; Kling says 10-bit 4K is coming soon. Use the official site, not a reseller.
-
Live
Seedance / Dreamina
ByteDanceSeedance 2.5 on DreaminaFirst-party home for Seedance is Dreamina (CapCut). 2.5 launch note: up to 30s and 50 references. Also reaches some CapCut timelines; check the market.
-
Live
Veo
Google DeepMindVeo 3.1Official model page. Access is through Google’s own tools: Flow, Gemini, AI Studio and Vertex, not a third-party wrapper.
-
Live
Grok Imagine
xAIImagine image & videoOfficial access is Grok. Imagine does text-to-image and text-to-video in one thread, including reference-photo edits. Product note lists up to 15-second video.
-
Live
Runway
RunwayGen-4.5 and suiteOfficial app and API. Gen-4.5 is the current first-party video line; the docs list what subscribers and the API can actually call.
-
Live
Hailuo
MiniMaxHailuo videoMiniMax’s video line (Hailuo / H3). Official product door is hailuoai.video. Check the current Hailuo tier there; reseller catalogues lag the official name.
-
Live
Wan
AlibabaTongyi WanxiangAlibaba’s first-party Wanxiang / Wan entry. Open weights also circulate, this row is the official product door, not a mirror.
-
Live
Ray
LumaRay 3.2Official Luma Ray door. Ray 3.2 is the live video line (text/image-to-video, Modify Video, Reframe). Confirm current Ray version on Luma before you quote a spec.
-
Live
LTX Studio
LightricksLTX-2.5Official LTX Studio. LTX-2.5 is the live first-party video line on Studio; open LTX weights also exist elsewhere. This row is the first-party app.
Voice over
Narration, dialogue and dubbing-
Live
ElevenLabs
ElevenLabsEleven v4 and v4 TurboOfficial ElevenLabs door for narration, character voices, dubbing and voice design. Current models are Eleven v4, with v4 Turbo for low-latency work. The voice over lane, not a video model.
-
Live
OpenAI Text to Speech
OpenAIRealtime API voicesOfficial OpenAI audio guide. The Speech API models tts-1, tts-1-hd and gpt-4o-mini-tts are deprecated and will shut down on 6 January 2027. The stated successor is gpt-realtime-2.1-mini, used through the Realtime API, not as a drop-in for the Speech API.
-
Live
Gemini TTS
GoogleGemini 3.8 Flash TTSOfficial Google Gemini text-to-speech door. The current flagship is Gemini 3.8 Flash TTS, with Gemini 3.8 Flash-Lite TTS as the faster, lower-cost sibling. Access is through Google AI Studio and the Gemini API.
-
Live
Fish Audio Drama 3
Fish AudioDrama 3Official Fish Audio Drama 3 door for directed, emotive text to speech, including multi-speaker reads. Open weights are under the Fish Audio Community Licence: free commercial use while yearly revenue is under US$500,000, with a separate deal needed above that or to host it as a voice service.
-
Live
Inworld
InworldRealtime TTS-2Official Inworld Realtime TTS-2 door for text to speech and voice cloning. TTS-2 is the flagship model; Realtime TTS-2 Flash is the faster, lower-cost sibling.
-
Live
MiniMax Speech
MiniMaxSpeech 2.8Official MiniMax Speech 2.8 door for text to speech. The API offers speech-2.8-hd for higher quality and speech-2.8-turbo for faster generations, with emotion and sound tags.
Music
Music and sound generation-
Live
Suno
SunoSuno v6Official Suno v6 door for original songs, score ideas and music sketches. Commercial use is included on paid plans; check the pricing page on Suno.
-
Live
Udio
UdioMusic generationOfficial Udio music generation door for song and soundtrack development. Check current access and terms on Udio.
-
Live
Stable Audio
Stability AIStable Audio 3.0Official Stable Audio 3.0 door for music and sound from text. You own the outputs under Stability AI’s Community Licence, with Enterprise cover if your organisation’s annual revenue is over US$1 million. The models are aimed mainly at instrumental work, not intelligible vocals. The web app is at stableaudio.com.
-
Live
Lyria
Google DeepMindLyria 3.5Google DeepMind’s first-party Lyria 3.5 door for music generation. Access is through Google Flow Music, Gemini, Google AI Studio and the Gemini API.
-
Live
Mureka
MurekaMureka V9.5Official Mureka V9.5 door for full songs with vocals, lyrics and instrumentals. The maker’s site states you can download tracks with full commercial rights; confirm the current terms on Mureka. API access is through the Mureka API Platform at platform.mureka.ai.
-
Live
ElevenLabs Music
ElevenLabsMusic v2.5Official ElevenLabs Music v2.5 door for studio-grade songs from text prompts, including composition plans, audio reference and inpainting. This is the music lane, separate from the ElevenLabs voice tools.
-
Live
MiniMax Music
MiniMaxMusic 3.0Official MiniMax Music 3.0 door for complete songs from a concept and optional lyrics. From 20 August 2026 the paid music APIs have been closed to new users; existing paying users can stay. New users are pointed to MiniMax Audio, or to the open-weight MiniMax Music 3 model.
Platforms
Multi-model access hubs-
Live
Dreamina
ByteDanceMulti-model image & video hubOfficial ByteDance hub (CapCut login). About page lists Seedream 5.0 Pro and GPT Image 2 for stills, Seedance 2.5 and 2.0 for video. Seedance stays on the Visual row; this is the platform door.
-
Live
Magnific (formerly Freepik)
MagnificMulti-model image & video hubOfficial Magnific AI suite; Freepik rebranded as Magnific in April 2026. Their own image-model docs list several Flux lines in one generator. First-party door only, not a stock mirror or credit reseller.
-
Live
Krea
KreaMulti-model image & video hubOfficial Krea app. Compare mode runs more than one model from the same prompt (own Krea line plus partner doors such as Nano Banana, ChatGPT image and Luma). Platform door, not a model-maker row.
-
Live
OpenArt
OpenArtMulti-model image & video hubOfficial OpenArt hub. One account reaches several image and video doors (Seedream, Seedance, Kling, Veo and others on their model pages). Credit the maker of a find; this row is only the platform door.
-
Live
IPVISE
IPVISEMulti-model canvas with agentOfficial IPVISE canvas. By their own account it puts an agent, node-based workflow, image editing and rough cuts on one canvas, and pulls in third-party image and video models. Their site lists teams in the UK, Japan and China. We plan to use it but have not tested it yet, so this is a pointer, not a review.
Tools
Products from the makers themselvesVoice, dubbing and other creative AI tools, each linked to the maker's own page. We don't list directories or code repos here, just the people who built the thing.
-
Tool
Cartesia Sonic 3.6
CartesiavoicetoolsSonic 3.6 text to speech and Ink speech to text for low-latency voice pipelines, expressive voices and streaming.
-
Tool
Resemble AI
Resemble AIvoicefilmpodcasttoolsVoice creation and cloning with rapid or professional clones, voice design, multilingual speech and self-host options.
-
Tool
Speechify Studio
SpeechifyvoicefilmpodcasttoolsCreator studio for AI voiceovers, dubbing and audio video projects, with a broad voice and accent library.
-
ToolRespeecher
RespeechervoicefilmtoolsProfessional AI voice cloning and speech-to-speech for film, TV, games and high-end media.
New Models
Newly listed-
LiveGPT-Video
GPT-VideoClip editorGPT-Video turns a long podcast, interview or stream into vertical captioned clips, with the crop kept on the speaker. It picks clips from the transcript, shows why each one was chosen, and you change a clip by typing what you want, which it calls vibe editing. Every edit can be undone.
-
LiveChariot-1-TTS-Multilingual
ChariotText to speechChariot-1-TTS-Multilingual turns text into speech in English and 11 Indic languages, with more than 50 voices. It can mix two supported languages in one sentence, clone a voice from a few seconds of audio, and use context to read prices, dates and abbreviations aloud. Chariot documents a public API and an MCP server.
-
LiveTaan
TaanMusic for videoTaan scores a video by reading the timing of the cuts and the mood, and can also take a script, a reference track, or timed instructions in plain English. You can ask it to change one part, one instrument or one moment without generating a fresh track. It documents a public API and an MCP server, and a job may return separate instrument stems. Taan says the music is original, with perpetual rights for commercial use.
Nothing matches that.